A traffic state estimation method, system and device

Abstract

The application discloses a traffic state estimation method, system and device, and relates to the technical field of traffic state estimation. The method comprises the following steps: acquiring historical traffic flow data sets of a current traffic section; introducing fuzzy logic into a Poisson process model to model the arrival rate of vehicles at different sections; using a Markov chain to model the transition probability of traffic flow states, and introducing a dynamic transition matrix to capture the changes of traffic flow states at different time periods, so as to build a fuzzy state transition model based on the Markov chain; combining the fuzzy Poisson distribution model and the fuzzy state transition model based on the Markov chain to obtain a fuzzy joint probability density function, and estimating the traffic flow state; and the method can estimate the vehicle trajectory and the traffic state in various complex traffic scenarios by combining the two models.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic state estimation, in particular to a traffic state estimation method, system and device. BACKGROUND

[0002] With the acceleration of modern urbanization, traffic congestion problems are becoming increasingly serious, and urban traffic management and optimization are facing great challenges.

[0003] Traditional traffic state estimation methods, such as models based on classical traffic flow theory, often rely on idealized assumptions and are difficult to accurately reflect the complex dynamic characteristics of traffic. For example, based on fixed Poisson distribution model or Markov process, although it can effectively simulate the vehicle arrival rate and traffic state transition, it is difficult to fully deal with the complexity, ambiguity and uncertainty in the actual traffic flow. Specifically, the vehicle arrival rate, vehicle speed and other states fluctuate under different traffic conditions, and these fluctuations are usually random and ambiguous. The traditional Poisson distribution assumes that the vehicle arrival is completely random, and fails to fully consider the correlation between the time variation of the traffic flow and the road state; while the Markov model assumes that the state transition has a fixed transition probability matrix, which cannot effectively capture the ambiguity in the traffic state.

[0004] In summary, most traditional traffic state estimation methods are based on deterministic models, which usually assume that various aspects of traffic flow are predictable and stable. However, traffic flow in the real world often has high randomness and volatility. Many existing traffic state estimation methods fail to fully consider the randomness and uncertainty in traffic flow, making it difficult to accurately describe and predict the evolution of traffic flow. In complex dynamic traffic scenarios, they often show large errors. SUMMARY

[0005] In view of the deficiencies that the prior art fails to fully consider the randomness and uncertainty in traffic flow, and is difficult to accurately describe and predict the evolution of traffic flow, the present application proposes a traffic state estimation method, system and device, which builds a fuzzy Poisson distribution model by introducing fuzzy logic into the traditional Poisson process model, models the transition of traffic state using Markov process, and realizes traffic state estimation by combining the two models, thereby solving the problems existing in the prior art.

[0006] A traffic state estimation method, comprising the following steps:

[0007] Obtain a historical traffic flow data set of the current traffic section; the traffic flow data set includes vehicle arrival rate, vehicle position and speed;

[0008] According to the arrival rate of the vehicle, fuzzy logic is introduced into the Poisson process model to estimate the vehicle arrival rate in a fuzzy time interval, and a fuzzy Poisson distribution model is built; according to the position and speed of the vehicle in different time periods, a traffic flow state transition model is established by using a Markov chain, the transition probability between traffic flow states is adjusted by introducing fuzzy logic in the model, a dynamic transition matrix is introduced to capture the change of traffic flow state in different time periods, and a fuzzy state transition model based on Markov chain is built; by combining the fuzzy Poisson distribution model and the fuzzy state transition model based on Markov chain, a fuzzy joint probability density model of traffic flow state and vehicle arrival rate of the vehicle in a set time interval is constructed;

[0009] The vehicle arrival rate and the transition probability of the traffic flow state of the to-be-tested traffic section in the current time period are taken as the input of the fuzzy joint probability density model, and the probability distribution of the traffic flow state of the traffic section in a set time interval adjacent to the current time period is output, and the traffic state of the traffic section in the set time interval is estimated according to the probability distribution.

[0010] Further, after obtaining the historical traffic flow data set of the current traffic section, the traffic flow data set is preprocessed by using Bayesian inference, and the processing process includes the following steps:

[0011] The traffic flow data set is collected by a plurality of source sensors; the traffic flow state is set as , the prior probability is assumed to be , and the observation data is According to the Bayesian formula, the following is obtained:

[0012] ;

[0013] Among them, represents the vehicle arrival rate; represents the number of observed vehicles; represents the posterior distribution of the traffic flow state after the observation data is given; is an observation model, representing the probability of the observation data given ; is a prior distribution, is a normalization constant;

[0014] The data collected by each sensor is fused by using Bayesian inference, and for each sensor , the observation model is defined as:

[0015] ;

[0016] Wherein, represents the given , the sensor collected observation subject to the normal distribution with for mean, for variance, and the data collected by different sensors have different variances ;

[0017] The observation data of multiple sensors are jointly modeled and represented as:

[0018] ;

[0019] Wherein, is the observation data from different sensors;

[0020] Based on Bayesian inference, the posterior distribution of the fused traffic flow state is represented as:

[0021] ;

[0022] The maximum a posteriori probability estimation of the arrival rate is represented as:

[0023] ;

[0024] The noise in the traffic flow data is removed by Bayesian inference, and the specific process is as follows: define the relationship between the observation value and the true traffic flow state , represented as:

[0025] ;

[0026] Wherein represents Gaussian noise, and the observation model is introduced into the Bayesian formula to obtain:

[0027] ;

[0028] The noise is removed by the prior distribution , and the estimated value of the traffic state is dynamically adjusted.

[0029] Further, the fuzzy logic is introduced into the Poisson process model to estimate the vehicle arrival rate in the fuzzy time interval, and a fuzzy Poisson distribution model is built, which specifically includes the following steps:

[0030] Define the vehicle arrival rate in the time interval as:

[0031] ;​

[0032] in For time Vehicle arrival rate within the area; Vehicle arrival rate at the beginning of a specific time period; for Model state, For noise terms, It is a correction function based on external factors. As weight;

[0033] In time interval Number of vehicles arriving Satisfies a Poisson distribution:

[0034] ;

[0035] in Indicates time interval Arrival The probability of a vehicle; Represents time interval The number of vehicles arriving within the area; Indicates the number of vehicles;

[0036] The cumulative distribution function for vehicle arrivals is:

[0037] ;

[0038] in Indicates that the arriving vehicle is less than or equal to The cumulative probability;

[0039] The formula for superposition of multiple Poisson processes in different regions is:

[0040] ;

[0041] in This represents the total vehicle arrival rate, which is the sum of the arrival rates across multiple areas; For the first Vehicle arrival rate in each area; It refers to the number of regions;

[0042] The arrival probability distribution within each region is as follows:

[0043] ;

[0044] in For the first Arrival within the area The probability of a vehicle; For the first The number of vehicles arriving within each area; This represents the number of vehicles that have arrived.

[0045] Based on time interval The conditional probability distribution of the previous adjacent time period is:

[0046] ;

[0047] in Given the arrival rate of the previous time period, the time interval The probability of the number of vehicles arriving within the specified timeframe; The step size.

[0048] Furthermore, it also includes the vehicle arrival rate. Using fuzzy sets for state transition probabilities To represent its uncertainty; assumption and state transition probability It is a fuzzy number, and its membership functions are respectively For fuzzy numbers Its membership function is expressed as:

[0049] ;

[0050] in It is the center value of the arrival rate. It is the standard deviation;

[0051] Then in terms of vehicle arrival rate In the presence of ambiguity, the number of vehicle arrivals The probability distribution is expressed as:

[0052] ;

[0053] Among them, membership function The degree of fuzziness indicating vehicle arrival rate Indicates all Perform summation;

[0054] To further process the fuzzy Poisson distribution process, a fuzzy probability generation function is defined for the fuzzy Poisson distribution process. The fuzzy probability generation function is expressed as:

[0055] ;

[0056] Using the fuzzy probability generating function, the mean and variance of the fuzzy Poisson process are calculated, and expressed as follows:

[0057] ;

[0058] ;

[0059] in, The sum of the mean and the mean of a fuzzy Poisson process This represents the variance of the fuzzy Poisson process.

[0060] Furthermore, the construction process of the fuzzy state transition model based on Markov chains specifically includes the following steps:

[0061] state space Defined as the position and speed of a vehicle at any given moment: ;in For the first One state, including the vehicle's location. and speed ; For the vehicle in the The x-coordinate of each state; For the vehicle in the The ordinate of each state i 3; The vehicle is in the The velocity in each state;

[0062] State transition probability Represented as:

[0063] ;

[0064] in From state Transferred to The probability of; Is the speed from Transferred to The probability of;

[0065] Fuzzy state transition probability Assuming each element in the state transition matrix If is a fuzzy number, then the fuzzy state transition probability is expressed as:

[0066] .

[0067] Furthermore, the state transition matrix The construction includes the following steps:

[0068] Suppose there is a traffic state vector and state transition matrix Traffic conditions are characterized by The state transition matrix is ​​represented by... If it means: ,in They are the first The mean and variance of each feature;

[0069] State transition matrix It is a row-normalized matrix, where each element represents a state. Transition to state The probability of:

[0070] ;

[0071] in, It is a state and state The state between;

[0072] A deep neural network is used to learn the state transition matrix, where the calculation process for each layer is represented as follows:

[0073] ;

[0074] in It is the first Layer output; and They are the first Layer weight matrix and bias vector; It is the activation function; the final output of the output layer is represented as:

[0075] ;

[0076] in, This represents the state transition probability estimated by a deep neural network. For the output of the hidden layer of a deep neural network, W , b These are the weights and biases, respectively. For the current observation value Functions that perform nonlinear transformations Its weight;

[0077] Activation function Defined as:

[0078] ;

[0079] The cross-entropy loss function is used to measure the difference between the predicted state transition matrix and the true matrix, expressed as:

[0080] ;

[0081] in These are the actual state transition probabilities. It is the predicted state transition probability matrix;

[0082] Predicting the state transition matrix using a neural network According to the output of the neural network Construct a traffic state estimation model. Indicates from state i 3 to state The transition probability; assuming for the th If the state transition probabilities of the steps are predicted, then the probability distribution of this state can be expressed as:

[0083] ;

[0084] in It is the first The state transition probability matrix of the step. It is the probability distribution of the initial state.

[0085] Furthermore, the fuzzy joint probability density model is expressed as:

[0086] ;

[0087] in, Time interval t Fuzzy numbers representing the arrival rate of vehicles within the premises; Time interval t The probability of observing the state sequence within; The state transition probabilities are estimated using a neural network. For the hidden layer output of a neural network, the membership function The degree of fuzziness indicating vehicle arrival rate The fuzzy membership function represents the relationship between the state transition probability and the fuzzy membership function. express t The probability distribution of the state at time -1.

[0088] The present invention also includes a traffic state estimation system, comprising:

[0089] The acquisition module is used to acquire historical traffic flow datasets for the current traffic segment; the traffic flow datasets include vehicle arrival rate, vehicle location, and speed.

[0090] The probability density model construction module is used to estimate the vehicle arrival rate within a fuzzy time interval by introducing fuzzy logic into a Poisson process model based on the vehicle arrival rate, thus constructing a fuzzy Poisson distribution model. Based on the vehicle position and speed within different time periods, a traffic flow state transition model is established using Markov chains. Fuzzy logic is introduced into this model to adjust the transition probabilities between traffic flow states, and a dynamic transition matrix is ​​introduced to capture the changes in traffic flow states across different time periods, thus constructing a fuzzy state transition model based on Markov chains. By combining the fuzzy Poisson distribution model and the fuzzy state transition model based on Markov chains, a fuzzy joint probability density model of the traffic flow state and vehicle arrival rate within a set time interval is constructed.

[0091] The estimation module takes the vehicle arrival rate and traffic flow state transition probability of the traffic segment under test in the current time period as input to the fuzzy joint probability density model, outputs the probability distribution of the traffic flow state of the traffic segment in a set time interval adjacent to the current time period, and estimates the traffic state of the traffic segment in the set time interval based on the probability distribution.

[0092] The present invention also includes a traffic state estimation computer device, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the traffic state estimation method.

[0093] The present invention also includes a readable storage medium storing a computer program, the computer program including program instructions which, when executed by a processor, are used to perform the steps of the traffic state estimation method.

[0094] The traffic state estimation method provided by this invention has the following beneficial effects:

[0095] This invention addresses the uncertainty and fuzziness in traffic flow by introducing fuzzy logic into the traditional Poisson process model, estimating the number of vehicle arrivals within fuzzy time intervals. It uses Markov chains to model traffic state transitions, employs fuzzy logic to adjust the transition probabilities between traffic states, and introduces a dynamic transition matrix to capture changes in traffic flow states over different time periods, such as the probability of transitioning from a free-flowing state to a congested state, thus establishing a fuzzy state transition model based on Markov chains. By combining these two models, vehicle trajectories and traffic states can be estimated in various complex traffic scenarios. Attached Figure Description

[0096] Figure 1 This is a flowchart of traffic state estimation in an embodiment of the present invention;

[0097] Figure 2This is a flowchart of Bayesian inference in an embodiment of the present invention;

[0098] Figure 3 This is a simulation diagram of the Poisson distribution of vehicle arrivals in an embodiment of the present invention;

[0099] Figure 4 This is a simulation diagram of the state transition process in an embodiment of the present invention;

[0100] Figure 5 This is a flowchart of the fuzzy Poisson distribution modeling module in an embodiment of the present invention;

[0101] Figure 6 This is a fuzzy Poisson distribution map of vehicle arrivals in an embodiment of the present invention;

[0102] Figure 7 This is a heatmap of fuzzy inference rules in an embodiment of the present invention;

[0103] Figure 8 This is a flowchart of the state transition modeling module in an embodiment of the present invention;

[0104] Figure 9 This is a diagram of the predicted state transition matrix in an embodiment of the present invention;

[0105] Figure 10 This is a diagram showing the state transition matrix for different samples in an embodiment of the present invention; Figure 10 (a) is the state transition matrix of sample 1. Figure 10 (b) is the state transition matrix of sample 2. Figure 10 (c) is the state transition matrix of sample 3;

[0106] Figure 11 This is a comparison chart of the neural network predicted values ​​and the actual values ​​in an embodiment of the present invention;

[0107] Figure 12 This is a flowchart of the joint modeling process in an embodiment of the present invention;

[0108] Figure 13 This is a diagram showing the data and joint probability density distribution in an embodiment of the present invention; Figure 13 (a) is a graph of Poisson distribution data. Figure 13 (b) is a plot of fuzzy Poisson distribution data. Figure 13 (c) is the Markov chain state transition matrix. Figure 13 (d) is the radar diagram of the state probability distribution of the joint model;

[0109] Figure 14 This is a comparison chart of the results of Poisson distribution, fuzzy Poisson distribution, and Markov chain in the embodiments of the present invention; Figure 14 (a) is a schematic diagram of traffic state estimation using fuzzy Poisson distribution and Markov chain. Figure 14(b) is a schematic diagram of traffic state estimation using the Poisson distribution. Detailed Implementation

[0110] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0111] This invention proposes a traffic state estimation method, such as... Figure 1 As shown, the method specifically includes the following steps:

[0112] S1. Data Acquisition and Preprocessing Module: First, traffic flow data is collected through multi-source sensors (such as traffic cameras, radar, GPS, etc.), including raw data such as vehicle arrival time, vehicle speed, and road occupancy. For the collected discrete data, fuzzy logic is used to preprocess it to establish the basis for handling uncertainty and fuzziness.

[0113] This module incorporates Bayesian inference, dynamically updating traffic flow state estimates through multi-source data fusion and uncertainty handling. In this module, Bayesian inference is used to process information from different data sources (such as radar, cameras, GPS, etc.) and to address noise, uncertainty, and real-time requirements. Figure 2 As shown.

[0114] Definition 1: Bayesian inference: Bayesian inference is the ratio of posterior probability to prior probability, representing how likely it is that a hypothesis is true given observed data Z, relative to the case where a hypothesis is false.

[0115] Set traffic flow status as Perform a Bayesian estimation on it, assuming the prior probability is... The observation data is According to Bayes' theorem:

[0116] (1);

[0117] in, Represents arrival rate, Represents the number of vehicles observed. Represents given observation data Afterwards, traffic flow status The posterior distribution, It is an observation model, representing a given The probability of subsequent observation data, It is the prior distribution, reflecting the... Prior knowledge, It is a normalization constant.

[0118] Data collected from different sensors is fused from multiple sources and treated as multiple observations. Each sensor's measurement data has different noise characteristics and sampling frequencies, therefore Bayesian inference is needed for multi-source data fusion. For each sensor... Define its observation model for:

[0119] ;

[0120] in, Indicates that in a given In this case, the sensor Collected observations Obey For the mean, The variance follows a normal distribution, and data from different sensors have different variances. This reflects the observation noise from different data sources. Observational data from multiple sensors can be jointly modeled using the following formula:

[0121] (2);

[0122] in, These are observation data from different sensors. It is the noise variance of each sensor.

[0123] Based on Bayesian inference, the merged traffic flow state The posterior distribution can be expressed as:

[0124] (3);

[0125] To obtain the arrival rate The estimated value needs to be calculated by determining its maximum a posteriori probability estimate. The formula is as follows:

[0126] (4);

[0127] Traffic state data inevitably contains noise during the acquisition process, especially under conditions of sensor hardware limitations and external environmental influences. However, Bayesian inference can effectively eliminate the impact of noise and improve data quality by dynamically adjusting the model state. Assuming that sensor noise follows a Gaussian distribution, Bayesian inference can handle uncertainty by constructing a probability distribution of the observed noise. Define the observed values... and actual traffic flow status The relationship is as follows:

[0128] (5);

[0129] in Represents Gaussian noise. To represent the noise variance, the observation model is introduced into Bayes' theorem:

[0130] (6);

[0131] Based on prior distribution It can effectively remove noise and dynamically adjust the estimated traffic conditions.

[0132] To further address uncertainty and fuzziness, Bayesian inference is combined with fuzzy logic, using fuzzy membership functions to refine prior distributions and observation models. Traffic conditions are then defined. Belongs to a fuzzy set Its fuzzy membership function is Using Bayesian inference to adjust the prior and posterior distributions based on membership, the new Bayesian inference formula is:

[0133] (7);

[0134] Among them, fuzzy membership degree It plays a role in correcting model uncertainties within the Bayesian framework, dynamically adjusting traffic state estimates based on real-time data.

[0135] S2, Fuzzy Poisson Distribution Modeling Module, such as Figure 5 As shown: In the Poisson distribution, the arrival rate of vehicles This is a critical state, reflecting the expected number of vehicles arriving per unit time. In traditional models... To determine the value, and in this invention, fuzzy logic is introduced to... The data is processed as fuzzy numbers, and its fuzziness is represented using membership functions. The number of vehicle arrivals within a fuzzy time interval is estimated using a fuzzy Poisson distribution, such as... Figure 3 As shown, a joint probability model of the fuzzy Poisson distribution is constructed.

[0136] Definition 1: A fuzzy Poisson process incorporates fuzzy logic into the traditional Poisson process model to address the uncertainty and fuzziness in traffic flow. This model can handle the inherent fuzziness in real-world traffic data, such as the fuzziness of vehicle arrival rates and time intervals, thus providing realistic results for traffic state estimation.

[0137] Because vehicle arrivals are discrete and random, these events can be described using a Poisson distribution. However, since arriving vehicles are affected by traffic flow disturbances, their uncertainty needs to be considered; that is, the states (usually event rates) are no longer precise numerical values ​​but fuzzy sets. The fuzzy Poisson distribution of discrete random variables is a probability distribution model that combines concepts from the traditional Poisson distribution and fuzzy set theory.

[0138] In a fuzzy Poisson distribution, the event rate is a fuzzy number, which can be represented as a set of probability density functions, each corresponding to a possible Poisson state value. In this way, for a given set of data observations, the likelihood of all possible states can be calculated, resulting in a more comprehensive probabilistic interpretation, such as... Figure 7 As shown.

[0139] Vehicle arrival rate within a defined time interval:

[0140] (8);

[0141] in For time Vehicle arrival rate within the area; Initial vehicle arrival rate; For autoregressive integral moving average model Model state, For noise terms, It is a correction function based on external factors (such as weather, time, etc.). Assign weights to them.

[0142] In time interval Number of vehicles arriving Satisfies a Poisson distribution:

[0143] (9);

[0144] in Indicates time Arrival The probability of a vehicle; Representing time The number of vehicles arriving within the area; This indicates the specific number of vehicles.

[0145] Cumulative distribution function of vehicle arrivals:

[0146] (10);

[0147] in Indicates that the arriving vehicle is less than or equal to The cumulative probability.

[0148] Furthermore, the formula for superimposing multiple Poisson processes in different regions is as follows:

[0149] (11);

[0150] in The total vehicle arrival rate is the sum of the arrival rates across multiple areas; For the first Vehicle arrival rate in each area; It refers to the number of regions.

[0151] The arrival probability distribution within each region is as follows:

[0152] (12);

[0153] in For the first Arrival within the area The probability of a vehicle; For the first The number of vehicles arriving within each area; This refers to the specific number of vehicles that arrived.

[0154] The conditional probability distribution based on the previous time period is:

[0155] (13);

[0156] in Given the arrival rate of the previous time period, this is the probability of the number of vehicles arriving in the current time period. The step size.

[0157] Since vehicle arrival is uncertain, and fuzzy logic allows for the introduction of uncertainty when dealing with traffic conditions, the arrival rate of vehicles... The state transition probability can be represented by fuzzy sets. To represent its uncertainty. Therefore, we assume that this module's... and state transition module It is a fuzzy number, and its membership functions are respectively The membership function can be a triangular, trapezoidal, or Gaussian function. For fuzzy numbers... Its membership function can be expressed as:

[0158] (14);

[0159] in It is the center value of the arrival rate. It is the standard deviation.

[0160] After fuzzification, the Poisson distribution can be represented as a fuzzy Poisson distribution, as shown in the following formula:

[0161] (15);

[0162] The formula represents the vehicle arrival rate. In the presence of ambiguity, the number of vehicle arrivals probability distribution, membership function The integral represents the degree of fuzziness of the vehicle arrival rate for all possible [locations / regions]. To perform summation, such as Figure 6 As shown.

[0163] To further process the fuzzy Poisson distribution process, a fuzzy probability generation function for the fuzzy Poisson distribution process is defined, as follows:

[0164] (16);

[0165] The mean and variance of a fuzzy Poisson process can be calculated using the fuzzy probability generation function, as shown in the following formula:

[0166] (17);

[0167] (18);

[0168] The two formulas above represent the mean and variance of the fuzzy Poisson process, respectively. The mean represents the expected number of vehicle arrivals after considering fuzziness, and the variance reflects the volatility of the number of arrivals.

[0169] S3, State Transition Modeling and Markov Process Module, such as Figure 8 As shown: Markov chains are used to model traffic state transitions, and the transition probabilities between states are adjusted using fuzzy logic. A dynamic transition matrix is ​​introduced to capture changes in traffic flow states over different time periods. Combining historical data with real-time traffic information, a fuzzy state transition model based on Markov chains is established, as follows: Figure 4 As shown.

[0170] state space Defined as the position and speed of a vehicle at a given moment:

[0171] (19);

[0172] in For the first One state, including the vehicle's location. and speed ; For the vehicle in the The x-coordinate of each state; For the vehicle in the The ordinate of each state i 3; The vehicle is in the The speed in each state.

[0173] State transition probability :

[0174] (20);

[0175] in From state Transferred to The probability of; Is the speed from Transferred to The probability of.

[0176] Fuzzy state transition probability Assuming each element in the state transition matrix It is a fuzzy number, and the fuzzy state transition probability can be expressed as:

[0177] (twenty one);

[0178] State transition matrix The structure is as follows:

[0179] The learning-based state transition matrix is ​​a dynamic matrix automatically generated by a machine learning model to describe the transition probabilities between states in a traffic system. The core of this method is to use complex machine learning algorithms (such as deep neural networks and random forests) to train on a large amount of historical traffic data, thereby generating a state transition matrix that accurately reflects the current traffic situation. This invention uses a deep learning neural network to learn the state transition matrix.

[0180] The data generation part is represented using random variables and probability distributions, assuming there is a traffic state vector. and state transition matrix Traffic conditions are characterized by The state transition matrix is ​​represented by... This can be represented as follows:

[0181] (twenty two);

[0182] in It is the first The mean and variance of each feature.

[0183] State transition matrix It is a row-normalized matrix, where each element represents a state. Transition to state The probability of:

[0184] (twenty three);

[0185] in, It is a state and state The state between.

[0186] Using a deep neural network to learn the state transition matrix, specifically, the calculation of each layer can be represented as:

[0187] (twenty four);

[0188] in It is the first Layer output; and They are the first Layer weight matrix and bias vector; It's an activation function; the most commonly used one is... The activation function used in this invention is... The function uses a neural network to estimate the state transition probabilities in a Markov process, and deep learning is used to enhance the complex relationships between states. The final output layer is as follows:

[0189] (25);

[0190] in, This represents the state transition probability estimated by a deep neural network. For the output of the hidden layer of a deep neural network, These are the weights and biases, respectively. For the current observation value Functions that perform nonlinear transformations Assign weights to them.

[0191] Among them, the definition The function is defined as follows:

[0192] (26);

[0193] The cross-entropy loss function is used to measure the difference between the predicted state transition matrix and the true matrix, as shown in the following formula:

[0194] (27);

[0195] in These are the actual state transition probabilities. It is the predicted state transition probability matrix, such as Figure 9 , Figure 11 As shown.

[0196] Finally, the state transition matrix is ​​predicted using a neural network prediction model. According to the output of the neural network Construct a traffic state estimation model; this matrix represents the traffic state estimation model. i 3 to state The transition probability. Assume it is necessary to perform... For step prediction, the probability distribution of the state can be calculated using the following formula:

[0197] (28);

[0198] in It is the first The state transition probability matrix of the step. This represents the probability distribution of the initial state. Thus, the output vector... This can be transformed into a matrix representing state transition probabilities, such as... Figure 10 The diagram shown represents the transition matrix for different sample states.

[0199] S4. Fuzzy Poisson Distribution and Markov Joint Probability Density Model Module: This module combines the fuzzy Poisson distribution with a Markov process model, proposing a traffic state estimation method based on a joint probability density function. By using the vehicle arrival rate generated by the fuzzy Poisson distribution as the input to the probability density model, and combining it with the state transition characteristics of the Markov process, it can not only effectively handle the randomness of vehicle arrivals, but also address uncertainty through fuzzy logic, capturing the randomness and state dependence of traffic flow, thereby providing more accurate traffic flow prediction results and improving the accuracy of traffic state estimation.

[0200] By combining fuzzy Poisson distribution and Markov process, a time-series vehicle model is constructed. The joint probability density function of the state and arrival rate, combined with the fuzzy Poisson distribution and the fuzzy state transition probability, yields the fuzzified joint probability density function, as shown in the following formula:

[0201] (29);

[0202] in, The fuzzy number representing the vehicle arrival rate within the time interval; For time The probability of observing the state sequence at any given time; The state transition probabilities estimated by a deep neural network are obtained for Module 2; The output of the hidden layer of the deep neural network represents the fuzzy vehicle arrival rate. and fuzzy state transition probability Below, the number of vehicles arriving and state transition The joint probability formula, obtained by integrating the fuzzy variables, reflects the uncertainty of the data.

[0203] like Figure 12 As shown, the Arabic numerals 1-9 refer to the order of relationships between nodes, such as the steps from "data preprocessing" to "Poisson distribution modeling" and then to "fuzzification processing." They are used to identify the process connections between steps in the flowchart. Figure 12 In the diagram, A through F represent: data preprocessing, Poisson distribution modeling, fuzzification, Markov chain modeling, joint model solving, and result visualization, respectively. The results are as follows: Figure 13 As shown.

[0204] The estimated values ​​of the Poisson distribution and the fuzzy Poisson distribution are compared with the true state (the assumed Poisson mean). Error calculation: The mean squared error is used to measure the deviation of the 'Poisson estimate' and the 'fuzzy Poisson + Markov chain estimate' from the true state. Error comparison: The estimation errors of the two models are output, and the superiority of the fuzzy Poisson + Markov model is illustrated by comparing the magnitude of the errors. Figure 14 The estimation curves of Poisson distribution and fuzzy Poisson + Markov chain are displayed to facilitate a visual comparison of their accuracy. The estimation results for Poisson distribution and fuzzy Poisson + Markov chain are plotted separately. The results are as follows: Fuzzy Poisson + Markov estimation error: 0.001600000000000003; Poisson distribution estimation error: 21.77777777777782. This demonstrates that the method proposed in this invention is superior to traditional methods.

[0205] Based on the same inventive concept, this invention also proposes a traffic state estimation system, comprising:

[0206] The acquisition module is used to obtain the historical traffic flow dataset for the current traffic segment; the traffic flow dataset includes vehicle arrival rate, vehicle location, and speed.

[0207] The probability density model construction module is used to estimate the vehicle arrival rate within a fuzzy time interval by introducing fuzzy logic into a Poisson process model based on the vehicle arrival rate, thus constructing a fuzzy Poisson distribution model. Based on the vehicle position and speed within different time periods, a traffic flow state transition model is established using Markov chains. Fuzzy logic is introduced into this model to adjust the transition probabilities between traffic flow states, and a dynamic transition matrix is ​​introduced to capture the changes in traffic flow states across different time periods, thus constructing a Markov chain-based fuzzy state transition model. By combining the fuzzy Poisson distribution model and the Markov chain-based fuzzy state transition model, a fuzzy joint probability density model of the traffic flow state and vehicle arrival rate within a set time interval is constructed.

[0208] The estimation module takes the vehicle arrival rate and traffic flow state transition probability of the traffic segment under test in the current time period as input to the fuzzy joint probability density model, outputs the probability distribution of the traffic flow state of the traffic segment in a set time interval adjacent to the current time period, and estimates the traffic state of the traffic segment in the set time interval based on the probability distribution.

[0209] The present invention also proposes a traffic state estimation computer device, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the traffic state estimation method.

[0210] The present invention also proposes a readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, are used to perform steps of a traffic state estimation method.

[0211] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A traffic state estimation method characterized by comprising: The method comprises the following steps: Obtain a historical traffic flow data set of a current traffic section; the traffic flow data set comprises vehicle arrival rate, vehicle position and speed; According to the arrival rate of the vehicle, fuzzy logic is introduced into the Poisson process model to estimate the vehicle arrival rate in a fuzzy time interval, and a fuzzy Poisson distribution model is built; according to the position and speed of the vehicle in different time periods, a traffic flow state transition model is established by using a Markov chain, the transition probability between traffic flow states is adjusted by introducing fuzzy logic in the model, a dynamic transition matrix is introduced to capture the change of the traffic flow state in different time periods, and a fuzzy state transition model based on the Markov chain is built; by combining the fuzzy Poisson distribution model and the fuzzy state transition model based on the Markov chain, a fuzzy joint probability density model of the traffic flow state and the vehicle arrival rate of the vehicle in a set time interval is constructed; the fuzzy logic is introduced into the Poisson process model to estimate the vehicle arrival rate in a fuzzy time interval, and the fuzzy Poisson distribution model is built, which comprises the following steps: Defining time intervals The inner vehicle arrival rate is: ； wherein is the time vehicle arrival rate within; is the vehicle arrival rate at the beginning of a specific time period; is model state, is a noise term, is a correction function based on external factors, is a weight; arriving at the vehicle number of vehicles arriving at the vehicle satisfies a poisson distribution: ； wherein denotes a time interval arriving within probability of a vehicle; denotes a time interval arriving within denotes a vehicle count; The cumulative distribution function of vehicle arrival is: ； wherein represents the cumulative probability of reaching the vehicle less than or equal to the time t. The superposition formula of multiple Poisson processes in different regions is: ； wherein denotes the total vehicle arrival rate, which is the sum of the arrival rates of the plurality of zones; is the vehicle arrival rate of the zone; is the number of zones; The arrival probability distribution in each region is: ； in For the first Arrival within the area The probability of a vehicle; For the first The number of vehicles arriving within each area; This represents the number of vehicles that have arrived. Based on time intervals The conditional probability distribution of the previous adjacent time period is: ； wherein is the probability of the number of vehicles arriving in the time interval given the arrival rate in the previous time period; is the step size; Arrival rate of vehicles and state transition probabilities are represented by fuzzy sets to represent their uncertainty; it is assumed and state transition probabilities are fuzzy numbers whose membership functions are , respectively, for fuzzy numbers whose membership functions are represented as: ； wherein is the central value of the arrival rate, is the standard deviation; Then the vehicle arrival rate In the case of ambiguity, the number of vehicles arriving The probability distribution is expressed as: ； where the membership function denotes the fuzziness of the vehicle arrival rate, denotes the summation over all vehicle arrival rates; In order to further process the fuzzy Poisson distribution process, the fuzzy probability generating function of the fuzzy Poisson distribution process is defined, and the fuzzy probability generating function is represented as: ； The mean and variance of the fuzzy Poisson process are calculated by using the fuzzy probability generating function, and are represented as: ； ； wherein, denotes the mean of the fuzzy Poisson process denotes the variance of the fuzzy Poisson process; The vehicle arrival rate and the transition probability of the traffic flow state of the to-be-tested traffic section in the current time period are taken as the input of the fuzzy joint probability density model, and the probability distribution of the traffic flow state of the traffic section in a set time interval adjacent to the current time period is output, and the traffic state of the traffic section in the set time interval is estimated according to the probability distribution.

2. The traffic state estimation method according to claim 1, wherein After obtaining the historical traffic flow data set of the current traffic section, the traffic flow data set is preprocessed by using Bayesian inference, and the processing process comprises the following steps: Traffic flow data set is collected by multi-source sensors; traffic flow state is set as , the prior probability is assumed as , the observation data is , and then the Bayesian formula is obtained: ； where, represents the vehicle arrival rate; represents the observed number of vehicles; denotes the given observation data the posterior traffic flow state the posterior distribution; is the observation model, representing the probability of the observation data given the posterior traffic flow state; is the prior distribution, is the normalization constant; Bayesian inference is used to perform multi-source data fusion on the data collected by each sensor. For each sensor... Define its observation model for: ； wherein, represents the observation value collected by the sensor under the given condition obeys the normal distribution with the mean and the variance , and the data collected by different sensors have different variances ; The observation data of multiple sensors are jointly modeled and represented as: ； wherein, are observation data from different sensors; Based on Bayesian inference, the merged traffic flow state The posterior distribution is represented as: ； arrival rate maximum a posteriori probability estimate is expressed as: ； The noise in the traffic flow data is removed by Bayesian inference, the specific process is as follows: the relationship between the observation value And the real traffic flow state Is expressed as: ； where Introducing the observation model into the Bayes formula, we get: ； by the prior distribution The noise is removed and the estimate of traffic conditions is dynamically adjusted.

3. The traffic state estimation method according to claim 2, wherein The building process of the fuzzy state transition model based on the Markov chain comprises the following steps: State space is defined as the position and velocity of the vehicle at any instant in time: ; where is the th state, containing the position and velocity of the vehicle; is the horizontal coordinate of the vehicle at the th state; is the vertical coordinate of the vehicle at the th state i 3; is the velocity of the vehicle at the th state; State transition probabilities is represented as: ； wherein is the probability of transitioning from state to ; is the probability of transitioning from velocity to ; Fuzzy state transition probabilities , assuming each element in the state transition matrix is a fuzzy number, the fuzzy state transition probabilities are expressed as: 。 4. The traffic state estimation method according to claim 3, wherein The state transition matrix is constructed including the steps of: Assume there is a traffic state vector and a state transition matrix The traffic state is represented by the feature and the state transition matrix is represented by Then: where are the mean and variance of the th feature, respectively. State transition matrix is a row-normalized matrix, where each element represents the probability of transitioning from state to state . ； wherein is a state and a state between states; A deep neural network is used to learn the state transition matrix, and the calculation process of each layer is represented as: ； wherein is the output of the layer; and are the weight matrix and bias vector of the layer, respectively; is the activation function; the output of the final output layer is represented as: ； wherein, denotes a state transition probability estimated by a deep neural network, is a deep neural network hidden layer output, W , b are weights and biases, respectively, is a function that performs a non-linear transformation on the current observation , is a weight thereof; Activation function is defined as: ； A cross-entropy loss function is used to measure the difference between the predicted state transition matrix and the real matrix, and is represented as: ； wherein is the true state transition probability, is the predicted state transition probability matrix; Predicting state transition matrices by neural networks , from the output of the neural network , a traffic state estimation model is constructed, denotes the transition probability from state i 3 to state ; assuming that the state transition probabilities for the first step are predicted, the state probability distribution is represented as: ； wherein is the first step state transition probability matrix, is the initial state probability distribution.

5. The traffic state estimation method according to claim 4, wherein The fuzzy joint probability density model is represented as: ； where, is the time interval t is the fuzzy number of the vehicle arrival rate within is the time interval t is the state sequence observation probability within is the state transition probability estimated by the neural network is the neural network hidden layer output, the membership function denotes the fuzziness of the vehicle arrival rate, denotes the fuzzy membership function related to the state transition probability, denotes t the probability distribution of the state at time -1.

6. A traffic state estimation system characterized by comprising: Comprise: The obtaining module is used for obtaining a historical traffic flow data set of a current traffic section; the traffic flow data set comprises vehicle arrival rate, vehicle position and speed; The probability density model construction module is configured to introduce fuzzy logic into a Poisson process model to estimate vehicle arrival rate in a fuzzy time interval according to the arrival rate of the vehicle, to build a fuzzy Poisson distribution model; to build a traffic flow state transition model by using a Markov chain according to the position and speed of the vehicle in different time periods, to introduce fuzzy logic into the model to adjust the transition probability between traffic flow states, to introduce a dynamic transition matrix to capture the change of the traffic flow state in different time periods, and to build a fuzzy state transition model based on the Markov chain; and to build a fuzzy joint probability density model of the traffic flow state and the vehicle arrival rate of the vehicle in a set time interval by combining the fuzzy Poisson distribution model and the fuzzy state transition model based on the Markov chain. Defining time intervals The inner vehicle arrival rate is: ； wherein is the vehicle arrival rate within a time is the vehicle arrival rate at the beginning of a specific time period; model state, is a noise term, is a correction function based on external factors, is a weight;​​ arriving at the vehicle number of vehicles arriving at the vehicle satisfies a Poisson distribution: ； wherein represents a time interval arriving within probability of a vehicle; represents a time interval arriving within represents a vehicle number; The cumulative distribution function of the vehicle arrival is: ； wherein represents the cumulative probability of reaching the vehicle less than or equal to ; and represents the region; The superposition formula of the multiple Poisson processes in different regions is: ； wherein denotes the total vehicle arrival rate, which is the sum of the arrival rates of the plurality of zones; is the vehicle arrival rate of the th zone; is the number of zones; The arrival probability distribution in each region is: ； wherein is the probability of a vehicle arriving in the th zone; is the number of vehicles arriving in the th zone; is the number of vehicles arriving in the th zone; Based on time intervals The conditional probability distribution of the previous adjacent time period is: ； wherein is the probability of the number of vehicles arriving in the time interval given the arrival rate in the previous time period; is the step size; Arrival rate of vehicles and state transition probabilities are represented by fuzzy sets to represent their uncertainty; it is assumed and state transition probabilities are fuzzy numbers whose membership functions are , respectively, for fuzzy numbers whose membership functions are represented as: ； wherein is the central value of the arrival rate, is the standard deviation; Then the vehicle arrival rate In the case of ambiguity, the number of vehicles arriving The probability distribution is expressed as: ； where the membership function denotes the fuzziness of the vehicle arrival rate, denotes the summation over all vehicle arrival rates; To further process the fuzzy Poisson distribution process, a fuzzy probability generating function of the fuzzy Poisson distribution process is defined, and the fuzzy probability generating function is represented as: ； The mean and variance of the fuzzy Poisson process are calculated by using the fuzzy probability generating function, and are represented as: ； ； wherein, denotes the mean of the fuzzy Poisson process denotes the variance of the fuzzy Poisson process; The estimation module is configured to take the vehicle arrival rate and the transition probability of the traffic flow state of a to-be-tested traffic section in a current time period as inputs of the fuzzy joint probability density model, to output the probability distribution of the traffic flow state of the traffic section in a set time interval adjacent to the current time period, and to estimate the traffic state of the traffic section in the set time interval according to the probability distribution.

7. A traffic state estimation computer device characterized by comprising: The computer program includes program instructions, and the program instructions are executed by the processor to implement the steps of the traffic state estimation method in any one of claims 1-5. The computer program includes program instructions, and the program instructions are executed by the processor to implement the steps of the traffic state estimation method in any one of claims 1-5.

8. A readable storage medium, characterized by, ​

Images