Expressway traffic flow speed prediction method considering motion state difference between vehicles and hysteresis effect

Abstract

The invention discloses an expressway traffic flow speed prediction method considering motion state difference between vehicles and a hysteresis effect. The specific prediction method comprises the following six steps: step 1, calculating the disorder of the traffic flow caused by a single vehicle; 2, road unit distribution is carried out on disorder generated by a single vehicle to traffic flow; 3, calculating the total disorder of all road units; 4, constructing a hysteresis effect model; 5, determining the hysteresis phase number and coefficient of the hysteresis effect model; and 6, predicting the traffic flow speed. According to the method, the real-time track information of the vehicle is collected and the traffic flow speed is predicted according to the intelligent degree of the expressway assisted by detection equipment such as an induction coil and a vehicle-mounted GPS, so that traffic control measures can be taken in time, the purpose of preventing and relieving congestion is achieved, the operation efficiency and safety of a traffic system are improved, and the traffic safety is improved. And theoretical basis and model support are provided for expressway traffic flow speed prediction and traffic flow organization and control.

Description

Technical Field

[0001] The invention relates to the field of highway traffic flow prediction, and in particular to a method for predicting highway traffic flow speed taking into account the motion state difference and hysteresis effect between vehicles. Background Art

[0002] In recent years, the continuous development of sensing technology and communication technology, and the increasing popularity of intelligent monitoring equipment have provided support for accurate and real-time acquisition of vehicle operation information. Based on the comprehensive collection of highway traffic operation data, traffic flow prediction is carried out according to historical traffic information and real-time conditions to determine the future trend of traffic operation status, which is of great significance for achieving accurate control of highway traffic conditions, timely prevention and rapid relief of congestion, and improving highway service levels and operating efficiency.

[0003] In previous traffic flow forecasting research, domestic and foreign studies have used time series analysis, probability statistics methods and machine learning algorithms to capture the uncertainty and complex characteristics of time series changes in traffic flow parameters to perform traffic state deduction modeling and conduct traffic flow forecasting in specific scenarios. For example, patent 202210277813.X obtains GPS trajectory data of all vehicles in the study area within one month, calculates the approximate value of traffic flow speed at intervals of 15 minutes, constructs a traffic pattern similarity graph and a multi-graph convolutional network, and outputs a time series of traffic flow speed prediction values; patent 201811264868.7 collects historical traffic flow speed data within 180 days, establishes and calibrates the prediction model in the Internet of Vehicles environment, and realizes short-term traffic flow speed prediction; "Short-Term Traffic Speed ​​Forecasting Using a Deep Learning Method Basedon Multitemporal Traffic Flow Volume" is based on the relationship between flow and speed in traffic flow theory, and proposes a deep learning method for predicting traffic speed based on traffic flow data; "AMulti-step Traffic SpeedForecasting Model Based on Graph Convolutional LSTM" proposes a deep learning model based on historical traffic flow speed data and network topology, adopts graph convolutional network (GCN) to process the spatial relationship characteristics of traffic flow speed, and uses LSTM to process temporal relationship characteristics, so as to realize the prediction of multi-step traffic flow speed.

[0004] However, existing studies have the following three problems in predicting traffic flow speed based on historical data of macro traffic flow parameters such as flow and density or traffic flow speed time series itself: First, existing studies ignore the impact of micro vehicle motion differences on traffic status. In actual traffic scenarios, the vehicle motion state will be affected by other nearby vehicles, and the change of local vehicle motion state will also affect the overall traffic flow, which results in changes in the average speed of traffic flow after a period of time. Therefore, the change of traffic flow speed is not only affected by the changes of other macro parameters such as flow and density, but also the result of the difference in micro vehicle motion; second, existing studies require a large amount of historical data for traffic flow speed feature extraction and model training; third, existing studies are based on historical traffic data to predict traffic flow speed modeling, which can better reflect the change law of traffic flow speed time series in specific scenarios and specific time periods, but in the face of special circumstances and incidental events, the prediction effect of these models will be significantly reduced. Therefore, it is necessary to further analyze the relationship between macro traffic flow speed changes and micro vehicle motion states, solve the problem of difficulty in model training and calibration due to insufficient historical data, and ensure the accuracy and universal applicability of traffic flow speed prediction methods in various scenarios.

[0005] Based on the above background, if the traffic flow speed is regarded as the dependent variable and the microscopic vehicle movement difference as the independent variable, and considering that the change of traffic flow speed is relatively continuous, the value of the dependent variable will be affected by the change of the independent variable and the value of the dependent variable itself in the previous period. This phenomenon that the dependent variable is affected by its own lag value and the lag value of the independent variable is usually described by the lag effect model in economics. Therefore, it is urgent to design a highway traffic flow speed prediction method that takes into account the difference in movement state between vehicles and the lag effect, establish a lag effect model to express the impact of the microscopic vehicle movement state on the macroscopic traffic flow speed, use the collected real-time traffic flow data as the model input, calculate and output the time series of traffic flow speed prediction values, so as to provide a theoretical basis and model support for highway traffic control. Summary of the invention

[0006] In view of the shortcomings of existing research, the technical problem to be solved by the present invention is to propose a highway traffic flow speed prediction method that takes into account the differences in motion states and hysteresis effects between vehicles, so as to take traffic control measures in a timely manner to achieve the purpose of preventing and alleviating congestion, thereby improving the operating efficiency and safety of the transportation system, and providing a theoretical basis and model support for highway traffic flow speed prediction and traffic flow organization and control.

[0007] The present invention is implemented by adopting the following technical solutions, which are described as follows:

[0008] 1. A method for predicting highway traffic flow speed considering the difference in motion states between vehicles and hysteresis effect, characterized in that it comprises the following steps:

[0009] Step 1: Calculate the disturbance caused by a single vehicle to the traffic flow

[0010] First, obtain the traffic flow density ρ, the speed and displacement of each vehicle in the area to be predicted for the highway traffic flow speed, where the length of the area to be predicted for the highway traffic flow speed is not less than 150 meters, and the width is all lanes in the one-way driving direction. The traffic flow density ρ is the average number of vehicles per kilometer in the area to be predicted for the highway traffic flow speed; calculate the traffic flow speed:

[0011]

[0012] is the traffic flow speed, i.e., the average speed of all vehicles in the area to be predicted for the freeway traffic flow speed; N is the total number of vehicles in the area to be predicted for the freeway traffic flow speed; i is the vehicle number; i = 1, 2, ..., N; v i is the speed of vehicle i;

[0013] Second, establish a plane rectangular coordinate system, with the intersection of the lane edge line close to the central isolation belt and the upstream edge line of the highway traffic flow speed prediction area as the origin, set the traffic flow speed direction as the positive direction of the x-axis, and the direction perpendicular to the traffic flow speed direction as the y-axis direction, where the positive direction of the y-axis is rotated 90 degrees counterclockwise along the positive direction of the x-axis;

[0014] Third, decompose the speed of each vehicle along the positive direction of the x-axis and the positive direction of the y-axis to obtain the component speed v along the direction of the traffic flow speed i cosθ i The component velocity v perpendicular to the direction of traffic flow velocity i sinθ i , where θ i is the angle between the speed direction of vehicle i and the speed direction of traffic flow, 0≤θ i ≤π;

[0015] Fourth, calculate the speed difference between each vehicle:

[0016] △v xij =v i cosθ i -v j cosθ j

[0017] △v xij is the velocity difference between vehicle i and vehicle j along the x direction; v j cosθ j is the velocity component of vehicle j along the positive direction of x-axis, i is the vehicle number; j = 1, 2, ..., N;

[0018] △v yij =vi sinθ i -v j sinθ j

[0019] △v yij is the velocity difference between vehicle i and vehicle j along the y direction; v j sinθ j is the component velocity of vehicle j along the positive direction of the y-axis;

[0020] Fifth, calculate the displacement difference between each vehicle:

[0021] △x ij =x i -x j

[0022] △x ij is the displacement abscissa difference between vehicle i and vehicle j; i is the displacement abscissa of vehicle i; j is the displacement abscissa of vehicle j;

[0023] △y ij =y i -y j

[0024] △y ij is the displacement ordinate difference between vehicle i and vehicle j; y i is the displacement ordinate of vehicle i; y j is the displacement ordinate of vehicle j;

[0025] 6. Calculate the average vehicle spacing

[0026]

[0027] L is the length of the highway traffic flow speed prediction area; M is the number of lanes;

[0028] Seventh, use turbulence to express the difference in motion state between vehicles in the traffic flow and calculate the turbulence E caused by a single vehicle to the traffic flow i :

[0029]

[0030] It is assumed that when the speed and direction of all vehicles in the traffic flow are the same, the disturbance caused by a single vehicle to the traffic flow is 0;

[0031] Step 2: Assign road units to the disturbance caused by a single vehicle to the traffic flow

[0032] The highway traffic flow speed prediction area is divided into several rectangular road units of equal size, and the disturbance E caused by vehicle i to the traffic flow is i Assign to each road unit and calculate the single vehicle disorder distribution value E of the road unit i,p,q :

[0033]

[0034] p is the horizontal coordinate of the geometric center of the road unit; q is the vertical coordinate of the geometric center of the road unit; L i,p,q is the distance from the geometric center of vehicle i to the geometric center of the road unit (p, q);

[0035] The single vehicle disorder allocation value of each vehicle-to-road unit is added up to calculate the total disorder allocation value E of the road unit. p,q ;

[0036]

[0037] Step 3: Calculate the total disorder of all road units

[0038] The total disorder distribution value of all road units is added up to calculate the total disorder E of the area to be predicted for the highway traffic flow speed. a :

[0039]

[0040] g is the road unit number, g = 1, 2, ..., G; G is the total number of road units;

[0041] Step 4: Constructing a hysteresis effect model

[0042] The hysteresis effect theory is used to express the functional relationship between the traffic flow speed and the total disorder of the highway traffic flow speed in the predicted area, and the hysteresis effect model is established:

[0043]

[0044] in and are the traffic flow speeds at time t, time t-1, time t-2 and time tm, respectively, where the time interval between time t and time t-1 is the duration of one lag period, and the duration of one lag period is not greater than 2 seconds and not less than 0.1 seconds; E a (tl), E a (tl-1), E a (tl-2) and E a(tl-2) are the total disorder of the highway traffic flow speed in the area to be predicted at the time of tl, tl-1, tl-2 and tls respectively; l is the initial lag period of the total disorder of the highway traffic flow speed in the area to be predicted, l∈N; s is the maximum lag period of the total disorder of the highway traffic flow speed in the area to be predicted, s∈N + ; m is the maximum lag period of traffic flow speed, m∈N + ; α is the constant term of the lag effect model; β c is the coefficient, c=0, 1, 2, …, s; γ d is the coefficient, d=1, 2,…, m; u(t) is the random error;

[0045] Step 5: Determine the lag period and coefficient of the lag effect model

[0046] The objective function is constructed using the Akaike information criterion:

[0047]

[0048] Among them, AIC is the objective function; SSR is the residual sum of squares; n is the sample size, that is, the total number of moments included in the period of the input model;

[0049] Calculate the objective function values ​​for different combinations of s, l, and m values, and obtain the optimal s, l, and m values ​​corresponding to the minimum value of the objective function AIC;

[0050] The stepwise regression method is used to estimate the constant term and coefficient of the lag effect model, and the α, β c , γ d and u(t);

[0051] Step 6: Predict traffic flow speed

[0052] 6-1) The total disorder of the highway traffic flow speed in the area to be predicted and the traffic flow speed are used as the input of the lag effect model, where the corresponding time of the first and last item of the total disorder time series of the highway traffic flow speed in the area to be predicted is tl period and tls period respectively, and the corresponding time of the first and last item of the traffic flow speed time series is t-1 period and tm period respectively; the predicted value of the traffic flow speed at the time of t+1 period is calculated as the model output, that is,

[0053] 6-2) As the new model input, update the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area and the traffic flow speed time series. Then the corresponding time of the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area are t-l+1 period and tl-s+1 period respectively, and the corresponding time of the first and last items of the traffic flow speed time series are t and t-m+1 period respectively; calculate the model output and get the traffic flow speed prediction value at t+2 period

[0054] 6-3) Repeat steps 6-1 and 6-2 to calculate the traffic flow speed at each future moment and obtain a time series of traffic flow speed prediction values.

[0055] Compared with the prior art, the highway traffic flow speed prediction method considering the motion state difference and hysteresis effect between vehicles according to the present invention has the following beneficial effects:

[0056] (1) The method defines the difference in motion state between microscopic vehicles as disorder, and realizes the quantitative expression of the disorder caused by a single vehicle to the traffic flow, the single vehicle disorder allocation value of the road unit, the total disorder allocation value of the road unit, and the total disorder of the highway traffic flow speed to be predicted area, and determines the impact of the change in the motion state of microscopic vehicles on the macroscopic traffic flow speed;

[0057] (2) The method establishes a lag effect model with traffic flow speed as the dependent variable and the total disorder of the highway traffic flow speed prediction area as the independent variable. It does not require a large amount of historical traffic data for model training or calibration. The model input is the collected real-time vehicle trajectory information, and the output is the time series of traffic flow speed prediction values, which reduces the restrictions on data sources and data volume.

[0058] (3) The method is suitable for predicting highway traffic flow speed under various traffic conditions. Compared with the prediction method based on historical data, it has good scenario transferability and higher accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is an overall flow chart of the highway traffic flow speed prediction method considering the motion state difference and hysteresis effect between vehicles according to the present invention;

[0060] Figure 2 is a schematic diagram of calculating the turbulence caused by a single vehicle to the traffic flow according to the present invention;

[0061] Figure 3 is a schematic diagram of a single vehicle disorder distribution value of a road unit according to the present invention;

[0062] Figure 4 is a schematic diagram of the total value of disorder distribution of the road unit according to the present invention;

[0063] Figure 5 It is a schematic diagram of predicted traffic flow speed according to the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the details of the present invention and the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] The present invention proposes a highway traffic flow speed prediction method that takes into account the difference in motion states and hysteresis effects between vehicles. In practical application, the real-time trajectory information of vehicles is collected by induction coils, vehicle-mounted GPS and other detection equipment according to the intelligence level of the highway. By estimating the changing trend of traffic flow speed, traffic control measures are taken in time to achieve the purpose of preventing and alleviating congestion, thereby improving the operating efficiency and safety of the transportation system, and providing a theoretical basis and model support for highway speed prediction and traffic flow organization and control.

[0066] See also Figure 1 The method for predicting the speed of highway traffic flow considering the difference in motion states between vehicles and the hysteresis effect described in the present invention comprises the following steps, which are discussed in detail below:

[0067] Step 1: Calculate the disturbance caused by a single vehicle to the traffic flow

[0068] First, obtain the traffic flow density ρ, the speed and displacement of each vehicle in the area to be predicted for the highway traffic flow speed, where the length of the area to be predicted for the highway traffic flow speed is not less than 150 meters, and the width is all lanes in the one-way driving direction. The traffic flow density ρ is the average number of vehicles per kilometer in the area to be predicted for the highway traffic flow speed; calculate the traffic flow speed:

[0069]

[0070] is the traffic flow speed, i.e., the average speed of all vehicles in the area to be predicted for the freeway traffic flow speed; N is the total number of vehicles in the area to be predicted for the freeway traffic flow speed; i is the vehicle number; i = 1, 2, ..., N; v i is the speed of vehicle i;

[0071] Second, establish a plane rectangular coordinate system, with the intersection of the lane edge line close to the central isolation belt and the upstream edge line of the highway traffic flow speed prediction area as the origin, set the traffic flow speed direction as the positive direction of the x-axis, and the direction perpendicular to the traffic flow speed direction as the y-axis direction, where the positive direction of the y-axis is rotated 90 degrees counterclockwise along the positive direction of the x-axis;

[0072] Third, decompose the speed of each vehicle along the positive direction of the x-axis and the positive direction of the y-axis to obtain the component speed v along the direction of the traffic flow speed i cosθ i The component velocity v perpendicular to the direction of traffic flow velocity i sinθ i , where θ i is the angle between the speed direction of vehicle i and the speed direction of traffic flow, 0≤θ i ≤π;

[0073] Fourth, calculate the speed difference between each vehicle:

[0074] △v xij =v i cosθ i -v j cosθ j

[0075] △v xij is the velocity difference between vehicle i and vehicle j along the x direction; v j cosθ j is the velocity component of vehicle j along the positive direction of x-axis, i is the vehicle number; j = 1, 2, ..., N;

[0076] △v yij =v i sinθ i -v j sinθ j

[0077] △v yij is the velocity difference between vehicle i and vehicle j along the y direction; v j sinθ j is the component velocity of vehicle j along the positive direction of the y-axis;

[0078] Fifth, calculate the displacement difference between each vehicle:

[0079] △x ij =x i -x j

[0080] △x ij is the displacement abscissa difference between vehicle i and vehicle j; i is the displacement abscissa of vehicle i;j is the displacement abscissa of vehicle j;

[0081] △y ij =y i -y j

[0082] △y ij is the displacement ordinate difference between vehicle i and vehicle j; y i is the displacement ordinate of vehicle i; y j is the displacement ordinate of vehicle j;

[0083] 6. Calculate the average vehicle spacing

[0084]

[0085] L is the length of the highway traffic flow speed prediction area; M is the number of lanes;

[0086] Seventh, use turbulence to express the difference in motion state between vehicles in the traffic flow and calculate the turbulence E caused by a single vehicle to the traffic flow i :

[0087]

[0088] It is assumed that when the speed and direction of all vehicles in the traffic flow are the same, the disturbance caused by a single vehicle to the traffic flow is 0;

[0089] Step 2: Assign road units to the disturbance caused by a single vehicle to the traffic flow

[0090] The highway traffic flow speed prediction area is divided into several rectangular road units of equal size, and the disturbance E caused by vehicle i to the traffic flow is i Assign to each road unit and calculate the single vehicle disorder distribution value E of the road unit i,p,q :

[0091]

[0092] p is the horizontal coordinate of the geometric center of the road unit; q is the vertical coordinate of the geometric center of the road unit; L i,p,q is the distance from the geometric center of vehicle i to the geometric center of the road unit (p, q);

[0093] The single vehicle disorder allocation value of each vehicle-to-road unit is added up to calculate the total disorder allocation value E of the road unit. p,q ;

[0094]

[0095] Step 3: Calculate the total disorder of all road units

[0096] The total disorder distribution value of all road units is added up to calculate the total disorder E of the area to be predicted for the highway traffic flow speed. a :

[0097]

[0098] g is the road unit number, g = 1, 2, ..., G; G is the total number of road units;

[0099] Step 4: Constructing a hysteresis effect model

[0100] The hysteresis effect theory is used to express the functional relationship between the traffic flow speed and the total disorder of the highway traffic flow speed in the predicted area, and the hysteresis effect model is established:

[0101]

[0102] in and are the traffic flow speeds at time t, time t-1, time t-2 and time tm, respectively, where the time interval between time t and time t-1 is the duration of one lag period, and the duration of one lag period is not greater than 2 seconds and not less than 0.1 seconds; E a (tl), E a (tl-1), E a (tl-2) and E a (tl-2) are the total disorder of the highway traffic flow speed in the area to be predicted at the time of tl, tl-1, tl-2 and tls respectively; l is the initial lag period of the total disorder of the highway traffic flow speed in the area to be predicted, l∈N; s+l is the maximum lag period of the total disorder of the highway traffic flow speed in the area to be predicted, s∈N + ; m is the maximum lag period of traffic flow speed, m∈N + ; α is the constant term of the lag effect model; β c is the coefficient, c=0, 1, 2, …, s; γ d is the coefficient, d=1, 2,…, m; u(t) is the random error;

[0103] Step 5: Determine the lag period and coefficient of the lag effect model

[0104] The objective function is constructed using the Akaike information criterion:

[0105]

[0106] Among them, AIC is the objective function; SSR is the residual sum of squares; n is the sample size, that is, the total number of moments included in the period of the input model;

[0107] Calculate the objective function values ​​for different combinations of s, l, and m values, and obtain the optimal s, l, and m values ​​corresponding to the minimum value of the objective function AIC;

[0108] The stepwise regression method is used to estimate the constant term and coefficient of the lag effect model, and the α, β c , γ d and u(t);

[0109] Step 6: Predict traffic flow speed

[0110] 6-1) The total disorder of the highway traffic flow speed in the area to be predicted and the traffic flow speed are used as the input of the lag effect model, where the corresponding time of the first and last item of the total disorder time series of the highway traffic flow speed in the area to be predicted is tl period and tls period respectively, and the corresponding time of the first and last item of the traffic flow speed time series is t-1 period and tm period respectively; the predicted value of the traffic flow speed at the time of t+1 period is calculated as the model output, that is,

[0111] 6-2) As the new model input, update the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area and the traffic flow speed time series. Then the corresponding time of the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area are t-l+1 period and tl-s+1 period respectively, and the corresponding time of the first and last items of the traffic flow speed time series are t and t-m+1 period respectively; calculate the model output and get the traffic flow speed prediction value at t+2 period

[0112] 6-3) Repeat steps 6-1 and 6-2 to calculate the traffic flow speed at each future moment and obtain a time series of traffic flow speed prediction values.

[0113] Example

[0114] The following provides an implementation process and test results for a highway traffic flow speed prediction method that takes into account the motion state differences between vehicles and the hysteresis effect described in the present invention, but the protection scope of the present invention is not limited to the following embodiments.

[0115] The present invention selects the Beijing-Harbin Expressway from Shenyang to Siping, Liaoning Province for example analysis. The main line of the expressway is designed to have a speed of 120 km / h, with 8 lanes in both directions, and is equipped with conditions for obtaining vehicle operation information. The hourly distribution of the daily average actual traffic volume is used as the traffic operation condition to construct a simulation scenario.

[0116] The length of the highway traffic flow speed prediction area is set to 300 meters, the driving direction is upward, and the length of one lag period is 1.0 second. According to steps 1 to 3 of the method, the traffic flow speed at each moment, the disorder caused by a single vehicle to the traffic flow, and the total disorder of all road units are calculated with a step length of 1.0 second, and the traffic flow speed time series and the total disorder time series with a time span of 3600 seconds and a length of 7200 seconds are obtained.

[0117] According to steps 4 and 5 of the method, the lag effect model is constructed and the lag period and coefficient are calculated. The results are shown in Table 1.

[0118] Table 1. Lag period and coefficients of lag effect model

[0119]

[0120]

[0121] From Table 1, the initial lag number and the maximum lag number of the total disorder of the highway traffic flow speed in the predicted area are 6 and 11 respectively, the maximum lag number of the traffic flow speed is 2, and the duration of one lag is 1.0 second, indicating that in this scenario, the traffic flow speed value at the current moment is not only affected by the traffic flow speed value 1.0 second ago and 2.0 seconds ago, but also by the total disorder 6.0 seconds ago, 7.0 seconds ago, 8.0 seconds ago, 9.0 seconds ago, 10.0 seconds ago, and 11.0 seconds ago. In addition, the coefficients of each parameter in Table 1 are the degree of influence of the total disorder and traffic flow speed at different times on the traffic flow speed at the current moment.

[0122] According to step six of the method, the first 20 minutes of the traffic flow speed time series and the first 20 minutes of the total disorder time series are used as inputs to the lag effect model, and the traffic flow speeds at 20 minutes and 01 seconds, 20 minutes and 02 seconds, ... 59 minutes and 59 seconds are iteratively calculated with a step size of 1.0 second, and a time series 1 of traffic flow speed prediction values ​​with a total length of 2399 seconds is output.

[0123] The time series 1 of the predicted traffic flow speed is compared with the time series of the actual value of the traffic flow speed output by the simulation (the last 40 minutes), and the predicted root mean square percentage error is 3.47%, which proves the effectiveness of the method. In addition, in order to verify that the method has a good prediction effect without a large amount of historical data, the autoregressive integrated moving average model (ARIMA) is used to predict the traffic flow speed in this scenario. Time is used as the independent variable and traffic flow speed is used as the dependent variable. After passing the stationarity and ADF tests, the traffic flow speed time series of the first 20 minutes is also input with a step size of 1.0 second for model training, and the time series 2 of the predicted traffic flow speed of the last 40 minutes is output. The two time series of traffic flow speed prediction values ​​are compared with the actual values, and the distribution results of the absolute value of the error are shown in Table 2.

[0124] Table 2 Distribution of absolute value of errors in the time series of two traffic flow speed prediction values

[0125]

[0126] From Table 2, for the time series 1 of the traffic flow speed prediction value obtained by the method, the time when the absolute value of the error between the predicted value and the true value is [0,0.5) m / s accounts for 26.7% of the time series 1, and the time when the absolute value of the error is [0,1) m / s accounts for 73.2% of the entire time series; while for the time series 2 of the traffic flow speed prediction value obtained by ARIMA, the time when the absolute value of the error between the predicted value and the true value is [0,0.5) m / s accounts for 9.2% of the time series 2, and the time when the absolute value of the error is [0,1) m / s accounts for 46.4% of the entire time series. This shows that the prediction accuracy of the method is higher than that of the traffic flow speed prediction method based on historical data.

Claims

1. A highway traffic flow speed prediction method considering the difference in motion states between vehicles and hysteresis effect, characterized in that: The following steps are involved: Step 1: Calculate the disturbance caused by a single vehicle to the traffic flow First, obtain the traffic flow density ρ, the speed and displacement of each vehicle in the area to be predicted for the highway traffic flow speed, where the length of the area to be predicted for the highway traffic flow speed is not less than 150 meters, and the width is all lanes in the one-way driving direction. The traffic flow density ρ is the average number of vehicles per kilometer in the area to be predicted for the highway traffic flow speed; calculate the traffic flow speed: is the traffic flow speed, i.e., the average speed of all vehicles in the area to be predicted for the freeway traffic flow speed; N is the total number of vehicles in the area to be predicted for the freeway traffic flow speed; i is the vehicle number; i = 1, 2, ..., N; v i is the speed of vehicle i; Second, establish a plane rectangular coordinate system, with the intersection of the lane edge line close to the central isolation belt and the upstream edge line of the highway traffic flow speed prediction area as the origin, set the traffic flow speed direction as the positive direction of the x-axis, and the direction perpendicular to the traffic flow speed direction as the y-axis direction, where the positive direction of the y-axis is rotated 90 degrees counterclockwise along the positive direction of the x-axis; Third, decompose the speed of each vehicle along the positive direction of the x-axis and the positive direction of the y-axis to obtain the component speed v along the direction of traffic flow speed i cosθ i The component velocity v perpendicular to the direction of traffic flow velocity i sinθ i , where θ i is the angle between the speed direction of vehicle i and the speed direction of traffic flow, 0≤θ i ≤π; Fourth, calculate the speed difference between each vehicle: △v xij =v i cosθ i -v j cosθ j △v xij is the velocity difference between vehicle i and vehicle j along the x direction; v j cosθ j is the velocity component of vehicle j along the positive direction of x-axis, i is the vehicle number; j = 1, 2, ..., N; △v yij =v i sinθ i -v j sinθ j △v yij is the velocity difference between vehicle i and vehicle j along the y direction; v j sinθ j is the component velocity of vehicle j along the positive direction of the y-axis; Fifth, calculate the displacement difference between each vehicle: △x ij =x i -x j △x ij is the displacement abscissa difference between vehicle i and vehicle j; i is the displacement abscissa of vehicle i; j is the displacement abscissa of vehicle j; △and ij =and i -and j △y ij is the displacement ordinate difference between vehicle i and vehicle j; y i is the displacement ordinate of vehicle i; y j is the displacement ordinate of vehicle j; Sixth, calculate the average vehicle spacing r: L is the length of the highway traffic flow speed prediction area; M is the number of lanes; Seventh, use turbulence to express the difference in motion state between vehicles in the traffic flow and calculate the turbulence E caused by a single vehicle to the traffic flow i : It is assumed that when the speed and direction of all vehicles in the traffic flow are the same, the disturbance caused by a single vehicle to the traffic flow is 0; Step 2: Assign road units to the disturbance caused by a single vehicle to the traffic flow The highway traffic flow speed prediction area is divided into several rectangular road units of equal size, and the disturbance E caused by vehicle i to the traffic flow is i Assign to each road unit and calculate the single vehicle disorder distribution value E of the road unit i,p,q : p is the horizontal coordinate of the geometric center of the road unit; q is the vertical coordinate of the geometric center of the road unit; L i,p,q is the distance from the geometric center of vehicle i to the geometric center of the road unit (p, q); The single vehicle disorder allocation value of each vehicle-to-road unit is added up to calculate the total disorder allocation value E of the road unit. p,q ; Step 3: Calculate the total disorder of all road units The total disorder distribution value of all road units is added up to calculate the total disorder E of the area to be predicted for the highway traffic flow speed. a : g is the road unit number, g = 1, 2, ..., G; G is the total number of road units; Step 4: Constructing a hysteresis effect model The hysteresis effect theory is used to express the functional relationship between the traffic flow speed and the total disorder of the highway traffic flow speed in the predicted area, and the hysteresis effect model is established: in and are the traffic flow speeds at time t, time t-1, time t-2 and time tm, respectively, where the time interval between time t and time t-1 is the duration of one lag period, and the duration of one lag period is not greater than 2 seconds and not less than 0.1 seconds; E a (tl), E a (tl-1), E a (tl-2) and E a (tl-2) are the total disorder of the highway traffic flow speed in the area to be predicted at the time of tl, tl-1, tl-2 and tls respectively; l is the initial lag period of the total disorder of the highway traffic flow speed in the area to be predicted, l∈N; s is the maximum lag period of the total disorder of the highway traffic flow speed in the area to be predicted, s∈N + ; m is the maximum lag period of traffic flow speed, m∈N + ; α is the constant term of the lag effect model; β c is the coefficient, c=0, 1, 2, …, s; γ d is the coefficient, d=1, 2,…, m; u(t) is the random error; Step 5: Determine the lag period and coefficient of the lag effect model The objective function is constructed using the Akaike information criterion: Among them, AIC is the objective function; SSR is the residual sum of squares; n is the sample size, that is, the total number of moments included in the period of the input model; Calculate the objective function values ​​for different combinations of s, l, and m values, and obtain the optimal s, l, and m values ​​corresponding to the minimum value of the objective function AIC; The stepwise regression method is used to estimate the constant term and coefficient of the lag effect model, and the α, β c , γ d and u(t); Step 6: Predict traffic flow speed 6-1) The total disorder of the highway traffic flow speed in the area to be predicted and the traffic flow speed are used as the input of the lag effect model, where the corresponding time of the first and last item of the total disorder time series of the highway traffic flow speed in the area to be predicted is tl period and tls period respectively, and the corresponding time of the first and last item of the traffic flow speed time series is t-1 period and tm period respectively; the predicted value of the traffic flow speed at the time of t+1 period is calculated as the model output, that is, 6-2) As the new model input, update the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area and the traffic flow speed time series. Then the corresponding time of the first and last items of the total disorder time series of the highway traffic flow speed to be predicted area are t-l+1 period and tl-s+1 period respectively, and the corresponding time of the first and last items of the traffic flow speed time series are t and t-m+1 period respectively; calculate the model output and get the traffic flow speed prediction value at t+2 period 6-3) Repeat steps 6-1 and 6-2 to calculate the traffic flow speed at each future moment and obtain a time series of traffic flow speed prediction values.

Images