Method for short-term traffic risk prediction of road sections using roadside observation data

Abstract

Disclosed is a method for short-term traffic risk prediction of road sections by using roadside observation data. The method includes the following steps: 1) vehicle trajectory data in the detection area is obtained by using roadside observation data; 2) according to the continuous driving trajectories in the detection area, the traffic flow indicators are counted, and the surrogate safety indicators between vehicles are calculated; 3) time to collision and deceleration are selected as identification indicators to identify conflict events with collision risk in the detection area; 4) traffic flow indicators and surrogate safety indicators within the set time before the occurrence of conflict events are extracted, and the feature screening of various extracted indicators is performed by using classification algorithms; 5) based on the selected feature indicators, the indicators with the highest importance ranking are selected as the input to build a short-term traffic risk prediction model, and the model training and testing are completed by using the identified conflict events; 6) the short-term traffic risk prediction model is used to predict the risk of road sections. The proposed method can improve the prediction accuracy rate of road sections.

Claims

1. A method for short-term traffic risk prediction of road sections using roadside observation data, including the following steps: 1) the vehicle trajectory in the detection area is obtained by the roadside observation data; the basic information of vehicles in a preset detection area is collected by roadside detection equipment; the basic information includes a timestamp, vehicle ID, vehicle position and speed; based on the basic information of the vehicle stored in the radar record, the position information and speed information of the vehicles are extracted frame by frame according to the vehicle ID, and finally the trajectory data of each vehicle in the detection area is obtained; 2) according to the continuous driving trajectory in the detection area, the traffic flow indicators are counted and the surrogate safety indicators between vehicles are calculated; the surrogate safety indicators include: deceleration, distance headway, time headway, time to collision, modified time to collision, and stopping distance; traffic flow indicators include: traffic flow, occupancy rate, vehicle speed, as well as congestion index and the number of lane changes; 3) time to collision and deceleration are selected as identification indicators to identify conflict events with collision risk in the detection area; 4) traffic flow indicators and surrogate safety indicators within a set time before the occurrence of conflict events are extracted, and classification algorithms are used to perform feature screening on various indicators; 5) based on the selected feature indicators, the indicators with the highest importance ranking are selected as input to build a short-term traffic risk prediction model, and the identified conflict events are used to complete the model training and testing; 6) with the support of the constructed short-term traffic risk prediction model, the calculated feature indicators of a certain road section are selected as input to predict the traffic risk.

2. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 1, wherein the detection area in step 1) is set according to the longitudinal sensing range of the roadside detection device.

3. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 1, in step 1), the position information and speed information of the vehicle are extracted frame by frame according to the vehicle ID, and the trajectory data of each vehicle in the detection area is obtained, as follows: based on the basic information of the vehicle, the position information and speed information of the vehicle are extracted frame by frame according to the vehicle ID, and the trajectory data of each vehicle in the detection area is obtained; the expression is as follows:
Trajectory.sub.it={x.sub.i,y.sub.i,v.sub.i,t,ID} among them: Trajectory.sub.it represents the trajectory data of vehicle i at time t; x.sub.i, y.sub.i represent the location information of vehicle i; v.sub.i represents the speed information of vehicle i; ID represents the identification code information of vehicle i; the trajectory association of the nearest neighbor algorithm is adopted to deal with missing or interrupted trajectory data; the moving mean filtering method is used to obtain a smooth vehicle trajectory.

4. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 3, the trajectory association of the nearest neighbor algorithm is adopted to deal with missing or interrupted trajectory data; To supplement the missing trajectory data, the formula is as follows: $\mathrm{if}\left\{\left(x_i,y_i,v_i\right)|t=t_0\right\}=\varnothing ,$ $\{\begin{array}{c}{x}_i=x_i\left(t_0+\Delta t\right)\\ y_i=y_i\left(t_0+\Delta t\right)\\ v_i=v_i\left(t_0+\Delta t\right)\end{array}$ the moving mean filtering method is used to obtain a smooth vehicle trajectory; the mean filtering of the trajectory data is as follows: $\{\begin{array}{c}=\left[x_i\left(t-\Delta t\right)+x_i\left(t\right)+x_i\left(t+\Delta t\right)\right]/3\\ =\left[y_i\left(t-\Delta t\right)+y_i\left(t\right)+y_i\left(t+\Delta t\right)\right]/3\\ {\hat{v}}_{\mathrm{.Math.}}=\left[v_i\left(t-\Delta t\right)+v_i\left(t\right)+v_i\left(t+\Delta t\right)\right]/3\end{array}$ among them: Δt represents the time interval of recording trajectory data, , represents the position information of vehicle i after filtering, {circumflex over (v)}.sub.l represents the speed information of vehicle i after filtering processing.

5. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 1, the step 3) discriminates conflict events with collision risk in the detection area, specifically as follows: when both the time to collision and deceleration meet the threshold requirements, the event is regarded as a conflict event; $\mathrm{Event}=\{\begin{array}{ccc}1,\mathrm{TTC}<\mathrm{ttc}& \mathrm{and}& \mathrm{Decal}<a_{1\%}\\ 0,& & \mathrm{else}\end{array}$ among them, Event takes 1 to indicate that the event is a conflict event, and Event takes 0 to indicate that the event is a non-conflict event; TTC and Decel respectively represent the time to collision and deceleration value, ttc represents the critical threshold of the time to collision, a.sub.1% represents the 1% quantile of statistical deceleration.

6. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 1, wherein the step 4) extracts traffic flow indicators and surrogate safety indicators within a set time before the occurrence of conflict events. The classification algorithm is used to perform feature screening of various indicators, as follows: for the identified conflict events, based on the timestamp, the traffic flow indicators and surrogate safety indicators within a set time before each conflict event are extracted, and the extracted various indicators are aggregated for collision risk prediction modeling, as follows: the traffic flow indicators and surrogate safety indicators within a set time before each conflict event occurs are extracted, and the data aggregation method is used to obtain the aggregation value of each indicator; the data aggregation method is to calculate the average value of a certain indicator of all vehicle targets in the detection area in 1 second, and then aggregate the average value over the set time length; based on the aggregation values of various traffic flow indicators and surrogate safety indicators, according to the contribution of various indicators to collision risk prediction, the random forest classification algorithm is used to sort various indicators by feature importance to complete the screening of feature indicators.

7. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 6, wherein the data aggregation formula in step 4) is as follows: $\mathrm{Indicator}=\left[{.Math.}_t^{t+T}\left(\frac{{.Math.}_{n=1}^N{p}_{t,n}}{N}\right)\right]*\frac{1}{T}$ among them, Indicator represents the type of aggregated indicators, T represents the length of time for aggregation, t represents the starting point of data aggregation, N represents the total number of aggregated indicators of a certain type, P.sub.t,n represents the nth value of a certain indicator.

8. The method for short-term traffic risk prediction of road sections using roadside observation data according to claim 6, in step 4), the traffic flow indicators and surrogate safety indicators within a set time before each conflict event are extracted, and the set time is 60 to 120 seconds before the conflict event occurs.

9. A method for short-term traffic risk prediction of road sections using roadside observation data according to claim 1, wherein the construction of a short-term traffic risk prediction model in step 5) is to select the first M feature indicators ranked by importance from screening results as input and then construct a short-term traffic risk prediction model based on support vector machines.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0019] FIG. 1 is a flowchart of a method according to an embodiment of the present invention;

[0020] FIG. 2 is a schematic diagram of data aggregation time points in an embodiment of the present invention.

DETAILED DESCRIPTION

[0021] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not used to limit the present invention.

[0022] The following describes technical solutions of this application with reference to the accompanying drawings.

[0023] As shown in FIG. 1, a method for short-term traffic risk prediction of road sections using roadside observation data includes the following steps:

[0024] 1) The trajectory data of vehicles in the detection area is obtained by using the roadside observation data;

[0025] The basic information of vehicles in the detection area is collected at a fixed frequency by roadside microwave radar or other detection equipment; the detection area mainly refers to the longitudinal sensing range of the microwave radar; the basic information includes a timestamp, vehicle ID, vehicle position and speed;

[0026] The fixed acquisition frequency is generally set to 1 second/frame, which can be changed as required; the detection area is the longitudinal sensing range of the microwave radar, generally 150 m˜200 m;

[0027] Based on the basic information of the vehicles stored in the radar records, the position information and speed information of the vehicles are extracted frame by frame according to the vehicle ID, and the trajectory data of each vehicle in the detection area is obtained;

Trajectory.sub.it={x.sub.i,y.sub.i,v.sub.i,t,ID}

among them: [0028] Trajectory.sub.it represents the trajectory data of vehicle i at time t; [0029] x.sub.i, y.sub.i represent the location information of vehicle i; [0030] v.sub.i represents the speed information of vehicle i; [0031] ID represents the identification code information of vehicle i;

[0032] The missing or interrupted trajectory data is processed by adopting the trajectory association of the nearest neighbor algorithm; A smooth vehicle trajectory is obtained by using the moving mean filtering method;

[0033] To supplement the missing trajectory data, the formula is as follows:

[00001]$\mathrm{if}\left\{\left(x_i,y_i,v_i\right)|t=t_0\right\}=\varnothing ,$$\{\begin{array}{c}{x}_i=x_i\left(t_0+\Delta t\right)\\ y_i=y_i\left(t_0+\Delta t\right)\\ v_i=v_i\left(t_0+\Delta t\right)\end{array}$

[0034] The mean filtering is performed on the trajectory data, and the formula is as follows:

[00002]$\{\begin{array}{c}=\left[x_i\left(t-\Delta t\right)+x_i\left(t\right)+x_i\left(t+\Delta t\right)\right]/3\\ =\left[y_i\left(t-\Delta t\right)+y_i\left(t\right)+y_i\left(t+\Delta t\right)\right]/3\\ {\hat{v}}_{\mathrm{.Math.}}=\left[v_i\left(t-\Delta t\right)+v_i\left(t\right)+v_i\left(t+\Delta t\right)\right]/3\end{array}$

among them: [0035] Δt represents the time interval of recording trajectory data, [0036] , represents the position information of vehicle i after filtering, [0037] {circumflex over (v)}.sub.l represents the speed information of vehicle i after filtering processing;

[0038] 2) According to the continuous driving trajectory data in the detection area, the traffic flow indicators are counted and the surrogate safety indicators between vehicles are calculated; the surrogate safety indicators include: deceleration, distance headway, time headway, time to collision, modified time to collision, and stopping distance;

[0039] Traffic flow indicators include: traffic flow, occupancy rate, vehicle speed, as well as congestion index and the number of lane changes;

[0040] 3) Time to collision and deceleration are selected as identification indicators to identify conflict events with collision risk in the detection area;

[0041] According to the calculated multiple surrogate safety indicators, the time to collision and deceleration are selected as the distinguishing indicators, combined with ‘logic and’ criteria, to identify the conflict events with collision risk in the detection area. Time to collision and deceleration are indicators used to calibrate Near-Crash Events. The ‘logic and’ criterion requires that when both the time to collision and deceleration indicators meet the threshold requirements, the event will be recognized as a conflict event;

[00003]$\mathrm{Event}=\{\begin{array}{ccc}1,\mathrm{TTC}<\mathrm{ttc}& \mathrm{and}& \mathrm{Decal}<a_{1\%}\\ 0,& & \mathrm{else}\end{array}$

Among them, Event takes 1 to indicate that the event is a conflict event, and Event takes 0 to indicate that the event is a non-conflict event;
TTC and Decel respectively represent the time to collision and deceleration value,
ttc represents the critical threshold of time to collision, a.sub.1% represents the 1% quantile of the statistical deceleration;

[0042] 4) Traditional traffic flow indicators and typical surrogate safety indicators within a set time before the occurrence of conflict events are extracted, and feature screening on various indicators is performed by using classification algorithms;

[0043] For conflicts that have been identified, based on the timestamp, the traditional traffic flow indicators and typical surrogate safety indicators within 1 to 2 minutes before the occurrence of each conflict event are extracted. These various indicators are used for collision risk prediction modeling through data aggregation. The 0 to 1 minute period before the occurrence of the conflict is the response stage of safety measures after the collision risk prediction.

[0044] For the traditional traffic flow indicators and typical surrogate safety indicators within 1-2 minutes before the occurrence of the conflict, the data aggregation method is used to obtain the aggregated value of each indicator;

[0045] The data aggregation method is to calculate the average value of a certain indicator of all targets in the detection area in 1 second, and then aggregate the average value over a time segment length of 1 minute. The data aggregation formula is as follows:

[00004]$\mathrm{Indicator}=\left[{.Math.}_t^{t+T}\left(\frac{{.Math.}_{n=1}^N{p}_{t,n}}{N}\right)\right]*\frac{1}{T}$

[0046] Among them, Indicator represents the type of aggregated indicators,

T represents the length of time for aggregation, which is 1 minute,
t represents the starting point of data aggregation, that is, 2 minutes before the occurrence of the conflict event,
N represents the total number of aggregated indicators of a certain type,
P.sub.t,n represents the nth value of a certain type of indicator;

[0047] Based on the aggregation values of various traffic flow indicators and surrogate safety indicators, according to the contribution of various indicators to collision risk prediction, the random forest classification algorithm is used to sort the importance of various indicators to complete the screening of feature indicators.

[0048] In this embodiment, the indicator ranked in the top 90% of importance can be selected;

[0049] 5) Based on the sorting of feature indicators, according to actual needs, the indicators are selected as input to build a short-term traffic risk prediction model, and the identified conflict events are used to complete the training and testing of the model;

[0050] According to the screening results of feature indicators, the first M feature indicators ranked by importance are selected as input, and a short-term traffic risk prediction model based on support vector machines is constructed.

[0051] The identified conflict events are randomly divided into a training set and a test set according to a certain proportion. The training set is used to train the constructed short-term traffic risk prediction model, and the test set is used to verify the prediction effect of the short-term traffic risk prediction model.

[0052] 6) Based on the constructed short-term traffic risk prediction model, the screening results of feature indicators of a certain section are used as input to perform risk prediction.