Expressway disaster event occurrence time inference method considering traffic time-delay effect

Abstract

The invention discloses a highway disaster event occurrence time inference method considering a traffic time-delay effect. The method comprises the following steps: collecting key information of an event site to form an emergency data sequence; the data center matches the emergency data sequence with a database to obtain the longitude and latitude of a disaster incident occurrence place, and determines the numbers of upstream and downstream ETC gantries; screening out a list of delayed vehicles caused by disaster events by comparing the vehicle passing records of the downstream ETC portal frame with historical vehicle passing data; constructing a vehicle feature vector set of the disaster destruction event section; vehicle characteristics are concentrated in the vehicle characteristic vector; the section features and the feature vectors of the historical transaction tracks of the vehicles are input into a model for estimating the occurrence time of the highway disaster events for training and reasoning to obtain the time when the vehicles arrive at an accident site, and weighted average is carried out to obtain the estimated occurrence time of the disaster events; and the estimated occurrence time of the disaster events is used as the estimated occurrence time of the disaster events. According to the invention, the information query and processing efficiency is improved, and the timeliness of emergency response is guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of highway traffic emergency management, and in particular to a method for estimating the occurrence time of highway disaster events taking into account traffic time lag effects. Background Art

[0002] As the main artery of transportation, highways undertake a large number of passenger and freight transportation tasks. Their safety and smooth operation are directly related to economic development and the public's travel experience. However, in the process of operation, highways will inevitably encounter various disasters, such as traffic accidents, road damage caused by natural disasters, etc. These events will not only cause traffic congestion, but may also cause casualties and property losses. Therefore, quickly and accurately determining the time of the disaster is crucial for efficient emergency rescue, reasonable traffic diversion, and scientific identification of accident responsibility. In this context, the time tracing technology of highway disaster events based on ETC (electronic toll collection system) came into being. Relying on advanced information technology and rich ETC data resources, it is expected to provide strong support for solving the above problems. However, at the current stage, the application and development of related technologies still face many challenges.

[0003] Limitations of traditional methods for locating and timing disaster events: Traditionally, the location of highway disaster events often relies on manual on-site reports or simple mileage post number markings. Manual reports are easily affected by complex on-site conditions and subjective factors of personnel, and there are problems with inaccurate information and delayed reporting. Relying solely on mileage post number positioning, it is difficult to accurately determine the specific geographical location and cannot provide accurate guidance for the rapid arrival of rescue forces. In terms of determining the time of occurrence of disaster events, most of them are based on the time of discovery or the time recorded by relevant personnel, which has large errors and cannot meet the requirements for time accuracy in subsequent accident responsibility determination, traffic flow analysis and other work.

[0004] Insufficient use of existing data: A large number of devices are deployed along the highways, generating massive amounts of data, such as vehicle traffic data recorded by the ETC system, toll station transaction data, service area data, etc. However, these data are often stored in a scattered manner, lacking effective integration and in-depth mining. ETC data is only used for toll settlement, and the vehicle driving trajectory, speed and other information contained therein are not fully utilized to assist in the analysis of disaster events. At the same time, the formats of different data systems are not unified, and there are differences in vehicle identification, time format and section identification, and the data consistency is poor, which leads to difficulties in comprehensive analysis and cannot form a comprehensive and accurate basis for event analysis.

[0005] Lack of accurate time traceability model and technology: Currently, the traceability technology for the occurrence time of highway disaster and damage events is relatively scarce. Some existing analysis methods are mostly based on experience or simple statistical models, and cannot accurately consider various factors during vehicle driving, such as the average speed difference in different sections, the real-time impact of traffic conditions on vehicle speed, the impact of vehicle types on driving speed, etc. In the face of complex highway networks and changing traffic conditions, it is difficult to accurately calculate the specific time when the disaster and damage event occurred, and it cannot provide scientific and reliable decision-making support for traffic management departments.

[0006] In summary, there are many deficiencies in the existing traceability technology for the occurrence time of highway disaster and damage events, and it is difficult to meet the actual needs of quickly and accurately determining the disaster and damage time. There is an urgent need for a new technical solution to improve the positioning accuracy, optimize the data processing process, build a more accurate time calculation model, and improve the data query efficiency, so as to effectively trace the occurrence time of highway disaster and damage events. Summary of the Invention

[0007] The purpose of the present invention is to provide a method for inferring the occurrence time of highway disaster and damage events considering traffic time-delay effects, and use ETC data and related technologies to optimize the management of highway disaster and damage events, and improve the traffic management level and emergency handling ability.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A method for inferring the occurrence time of highway disaster and damage events considering traffic time-delay effects, comprising the following steps:

[0010] Step 1: Collect key information at the event site to form an emergency event data sequence; transmit the collected emergency event data sequence x i to a data center equipped with a GIS system for processing and analyzing the received data.

[0011] Among them, the key information at the event site includes but is not limited to mileage information (l), reporting time (t), and event type (u). The emergency event data sequence x i =<l i ,t i ,u i >, where i is the number of the data sequence x, and u i is the event type information used to identify the nature of the event.

[0012] Furthermore, in Step 1, the data center stores the emergency event data sequence in the historical record database to support subsequent event analysis and emergency response.

[0013] Step 2: The data center transmits the received emergency event data sequence x iMatch with the database and convert the mileage stake number to the longitude and latitude of the location where the disaster damage event occurred Achieve precise positioning of the event location; and determine the upstream and downstream ETC gantry numbers G i ,G i+1 , calculate the distance γ3 between the incident location and the upstream ETC gantry of the trip through the GIS system;

[0014] Step 3, by comparing the vehicle passing records of the downstream ETC gantry with the historical vehicle passing data, screen out the list of vehicles delayed due to the disaster damage event;

[0015] Step 4, construct the vehicle feature vector set F of the disaster damage event section F = <f1, f2, …, f n >, where n is the total number of vehicles, and the feature vector of each vehicle is the own feature of vehicle i, specifically including the historical passing speed of vehicle i and the type feature c of vehicle i i ; S = <v A , d> is the section feature, specifically including the historical speed feature v of the corresponding section A , the traffic flow structure feature d of the corresponding section; and the historical trajectory data of vehicle i Historical trajectory data includes all ETC gantries passed by the vehicle and their transaction times.

[0016] Step 5, use the machine learning framework to construct a model for estimating the occurrence time of highway disaster damage events; input the feature vectors of the own features of vehicle i, the section features, and the historical transaction trajectory of vehicle i into the model for estimating the occurrence time of highway disaster damage events for training and inference, and obtain the time ΔT when vehicle i arrives at the incident location i , and perform weighted average to obtain the estimated occurrence time T of the disaster damage event T = T i + ΔT is the estimated occurrence time of the disaster damage event.

[0017] Furthermore, in step 3, by selecting vehicles whose entry time into the section does not differ from the first batch of delayed vehicles by more than 1 minute, construct a vehicle list p i is the license plate number of vehicle i, is the time when vehicle i passes through the upstream ETC gantry. According to p i extract the historical ETC gantry transaction records of vehicle i from the ETC transaction flow data to obtain the vehicle historical trajectory data set n is the total number of vehicles, is the jth trip of vehicle i, m is the total number of trajectories of vehicle i, is the trip j of vehicle i, and at t2, t2, …, t gPassing through G1, G2, …, G respectively g Gantry. Step 3 specifically includes the following steps:

[0018] Step 31: By comparing the vehicle passing records of the downstream ETC gantry with the historical vehicle passing data, screen out the vehicles delayed due to disaster damage events;

[0019] Specifically, select the vehicles whose time of entering the section differs from that of the first p delayed vehicles by no more than 1 minute to construct a vehicle list where p i is the license plate number of vehicle i, is the time when vehicle i passes through the upstream ETC gantry.

[0020] Step 32: According to the vehicle license plate number p i Extract the historical ETC gantry transaction records of vehicle i from the ETC transaction flow data and generate the ETC trajectory data for each of its trips;

[0021] Step 33: Specifically, since the ETC transaction data contains a unique trajectory identifier, sort and aggregate it by transaction time to form the ETC trajectory data of vehicle flow i in trip j

[0022]

[0023] where, is the ETC gantry number passed by vehicle i in trip j; is the time when vehicle i passes through each ETC gantry.

[0024] Step 34, after generating the ETC trajectory data, integrate the multiple trip data of vehicle i to form the complete historical trajectory data of the corresponding vehicle. The specific expression is as follows:

[0025]

[0026] where m is the total number of historical trips of vehicle i;

[0027] Step 35, the historical trajectory data of all vehicles constitutes a data set

[0028]

[0029] where n is the total number of vehicles.

[0030] Furthermore, step 4 specifically includes the following steps:

[0031] Step 41, construct the section historical speed feature v A , and the expression is as follows:

[0032] v A =(α1, α2, α3)T (4)

[0033] α1 = max(v1, v2, …, v n ) (5)

[0034] α1 = min(v1, v2, …, v n ) (6)

[0035]

[0036] Where α1 is the maximum value of the section driving speed, α2 is the minimum value of the section driving speed, α3 is the average driving speed of the vehicles in the section, and v1, v2, …, v n are the speeds of the vehicles in the section before the delay occurs. The historical passing speed reflects the driving performance of the vehicles on different sections in the past. By analyzing the historical speed characteristics of the vehicles, the driving habits and speed change ranges of the vehicles can be understood.

[0037] Step 42: Construct the characteristics of the historical passing speed of vehicle i to obtain the historical passing speed v B of vehicle i, and the expression is as follows

[0038]

[0039] Where is the average passing speed of vehicle i in the immediately adjacent upstream section (the previous section), is the average passing speed of vehicle i in the upstream section before the previous one (the two previous sections).

[0040] Step 43: Construct the type characteristics c i of the vehicle, and the expression is as follows:

[0041] c i = (γ1, γ2, γ3) T (9)

[0042] Where γ1 represents the vehicle type, γ2 represents the vehicle axle speed, and γ3 represents the distance between the accident point and the previous ETC gantry

[0043] There are differences in the driving speeds and habits of different types of vehicles, and these differences will affect the time when the vehicles reach the disaster-damaged location. Constructing vehicle characteristics can comprehensively consider these factors. The vehicle type characteristics cover information such as vehicle type and number of axles. The driving speeds and handling performances of large trucks and small passenger cars are different, and their impacts on the traffic flow are also different. Incorporating these characteristics into the time estimation model can improve the accuracy of tracing the occurrence time of the disaster-damaged event.

[0044] Step 44: Obtain the proportion of different types of vehicles within the same time slice in the same section to form the traffic flow structure characteristics, and use the corresponding traffic flow structure characteristics as the section characteristics.

[0045] d = (δ1, δ2, δ3) T (10)

[0046] Among them, δ1 represents the proportion of passenger cars in the traffic flow, δ2 represents the proportion of freight trucks in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

[0047] Specifically, the traffic flow structure has a significant impact on the driving time of vehicles within the section. By constructing section characteristics, these influencing factors can be incorporated into the vehicle arrival time estimation model. Different proportions of passenger cars, freight trucks, and special vehicles in the traffic flow structure characteristics will lead to differences in traffic flow characteristics, thereby affecting the overall vehicle speed. If the proportion of freight trucks in a certain section is high, since the driving speed of freight trucks is relatively slow, it will reduce the overall vehicle speed of this section and increase the vehicle passing time.

[0048] Furthermore, Step 5 specifically includes the following steps:

[0049] Step 51: Use a machine learning framework to construct a model for estimating the occurrence time of highway disaster damage events; input vehicle characteristics, section characteristics, and historical transaction trajectories into the model for training and inference to predict the time ΔT when vehicle i arrives at the accident location. i ,

[0050]

[0051]

[0052] S = {v A , d}

[0053]

[0054] Among them, I is the vehicle characteristic of vehicle i, S is the section characteristic, and Tr i is the historical transaction trajectory of vehicle i.

[0055] Step 52: Calculate the average time for the vehicles arriving at the accident location extracted during this period.

[0056]

[0057] Among them, ΔT is the time for the disaster-damaged vehicle to reach the accident location from the ETC gantry upstream of the journey, and ΔT i is the time for vehicle i to reach the accident location from the ETC gantry upstream of the journey, and n is the total number of vehicles within the calculation range.

[0058] Then the estimated occurrence time of the disaster damage event is

[0059] T = T i + ΔT (15)

[0060] Wherein, T is the estimated occurrence time of the disaster-damaged event, and T i is the time when the vehicle flow with the earliest delay passes through the upstream ETC gantry.

[0061] Furthermore, in step 5, the LightGBM 3.4.0 is selected in the machine learning framework to construct a model for estimating the occurrence time of highway disaster-damaged events.

[0062] The present invention adopts the above technical solutions and has the following technical advantages: 1. The positioning and time traceability accuracy are greatly improved: The present invention utilizes the mileage stake number and longitude and latitude coordinate mapping table, combined with the GIS system, to improve the positioning accuracy of the disaster-damaged event to the meter level. For example, in multiple simulated disaster-damaged event tests, the positioning error of the traditional method is on average several hundred meters, while the present invention can be accurate to the specific lane position. 2. The data is utilized fully and efficiently: The present invention standardizes and cleans the ETC gantry data and integrates multi-source data. By deeply mining ETC data, such as vehicle driving trajectories and speeds, and combining with other data for comprehensive analysis, a more comprehensive perspective can be provided for the analysis of disaster-damaged events. 3. The time traceability model is accurate and reliable: The LightGBM 4.3.0 vehicle arrival time estimation model constructed by the present invention comprehensively considers factors such as vehicle speed, type, and section characteristics, and the prediction accuracy is significantly improved. Brief Description of the Drawings

[0063] The following further describes the present invention in detail in conjunction with the drawings and specific embodiments;

[0064] Figure 1 is a schematic flow chart of a method for inferring the occurrence time of highway disaster-damaged events considering traffic time delay effects according to the present invention;

[0065] Figure 2 is a schematic diagram of ETC trajectory data extraction according to the present invention;

[0066] Figure 3 is a schematic flow chart of screening a list of vehicles delayed due to disaster-damaged events according to the present invention;

[0067] Figure 4 is a schematic flow chart of extracting section characteristics according to the present invention. Specific Embodiments

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0069] Such as Figures 1 to 4As shown in one of them, the present invention discloses a method for inferring the occurrence time of highway disaster-damaged events considering traffic time-delay effects, and adopts a system block diagram of a method for inferring the occurrence time of highway disaster-damaged events considering traffic time-delay effects as shown in Figure 1 The specific steps are as follows:

[0070] Step 1: Collect key information at the event site to form an emergency event data sequence x i =<l i ,t i ,u i >, where i is the number of the data sequence x, and u i is the event type information used to identify the nature of the event, such as traffic accidents, natural disasters or other emergencies. Transmit the collected emergency event data sequence x i to the data center equipped with a GIS system for processing and analyzing the received data.

[0071] Among them, the key information at the event site includes but is not limited to mileage information (l), reporting time (t), and event type (u). The present invention combines the vehicle information collected by the ETC system (such as license plate number, timestamp, mileage of the accident location, etc.) with the geographic information system (GIS) to quickly calculate the location where the disaster event occurred. This method realizes the automatic collection, processing and analysis of emergency event data through the system, avoiding the defects of relying on manual work and on-site investigation in the traditional method.

[0072] Furthermore, in step 1, the data center stores the emergency event data sequence in the historical record database to support subsequent event analysis and emergency response.

[0073] Step 2: The data center matches the received emergency event data sequence x i with the database, converts the mileage to the longitude and latitude of the location where the disaster-damaged event occurred to accurately locate the event location; and determine the upstream and downstream ETC gantry numbers G i ,G i+1 , and calculate the distance γ3 between the accident location and the upstream ETC gantry of the journey through the GIS system;

[0074] Furthermore, the ETC transaction data used in step 2 needs to be pre-cleaned to filter out abnormal data. The main abnormalities in ETC transaction data are errcode abnormalities, vehicle identification abnormalities, and vehicle type abnormalities. After generating the above ETC trajectory, the gantry transaction information and gantry transaction time in the trajectory data can be recombined pairwise to obtain ETC section data.

[0075] Step 3: By comparing the vehicle passing records of the downstream ETC gantry with the historical vehicle passing data, screen out the list of vehicles delayed due to the disaster-damaged event;

[0076] Further, in step 3, vehicles with an entry section time differing from that of the first batch of delayed vehicles by no more than 1 minute are selected to construct a vehicle list. p i is the license plate number of vehicle i. is the time when vehicle i passes through the upstream ETC gantry. According to p i Extract the historical ETC gantry transaction records of vehicle i from the ETC transaction flow data to obtain the vehicle historical trajectory dataset. n is the total number of vehicles. is the j-th trip of vehicle i, and m is the total number of trajectories of vehicle i. is the trip j of vehicle i and passes through G1, G2,..., G g at t2, t2,..., t respectively. g gantries.

[0077] In step 4, construct a set of vehicle feature vectors F = <f1, f2,..., f n > for the disaster-damaged event section, where n is the total number of vehicles, and the feature vector of each vehicle is the self-feature of vehicle i, specifically including the historical passing speed of vehicle i and the type feature c i of vehicle i; S = <v A , d> is the section feature, specifically including the historical speed feature v A of the corresponding section and the traffic flow structure feature d of the corresponding section; and the historical trajectory data of vehicle i includes all ETC gantries passed by the vehicle and their transaction times.

[0078] In step 5, use a machine learning framework to construct a model for estimating the occurrence time of highway disaster-damaged events; input the self-feature of vehicle i, the section feature, and the feature vector of the historical transaction trajectory of vehicle i into the model for estimating the occurrence time of highway disaster-damaged events for training and inference to obtain the time ΔT i when vehicle i arrives at the accident location, and perform weighted averaging to obtain the estimated occurrence time T of the disaster-damaged event, where T = T i +ΔT is the estimated occurrence time of the disaster-damaged event.

[0079] Further, step 3 specifically includes the following steps:

[0080] Step 31: By comparing the vehicle passing records of the downstream ETC gantry with the historical vehicle passing data, screen out the vehicles delayed due to the disaster-damaged event.

[0081] Specifically, select the vehicles with an entry section time differing from that of the first batch of delayed vehicles by no more than 1 minute to construct a vehicle list. where p i is the license plate number of vehicle i, and is the time when vehicle i passes through the upstream ETC gantry.

[0082] Step 32: Extract the historical ETC gantry transaction records of vehicle i from the ETC transaction flow data according to the vehicle license plate number p i and generate the ETC trajectory data for each trip of it;

[0083] Step 33: As Figure 2 shown, since the ETC transaction data contains a unique trajectory identifier, sort and aggregate it according to the transaction time to form the ETC trajectory data of vehicle flow i in trip j

[0084]

[0085] where, is the ETC gantry number passed by vehicle i in trip j; is the time when vehicle i passes through each ETC gantry.

[0086] Step 34: After generating the ETC trajectory data, integrate the multiple trip data of vehicle i to form the complete historical trajectory data of the corresponding vehicle. The specific expression is as follows:

[0087]

[0088] where m is the total number of historical trips of vehicle i;

[0089] Step 35: The historical trajectory data of all vehicles constitutes a data set

[0090]

[0091] where n is the total number of vehicles.

[0092] Furthermore, step 4 specifically includes the following steps:

[0093] Step 41: Construct the section historical speed feature v A , and the expression is as follows:

[0094] v A =(α1,α2,α3) T (4)

[0095] α1 = max(v1,v2,…,v n ) (5)

[0096] α1 = min(v1,v2,…,v n ) (6)

[0097]

[0098] Among them, α1 is the maximum value of the section driving speed, α2 is the minimum value of the section driving speed, α3 is the average driving speed of the vehicles in the section, and v1, v2, …, v n are the speeds of the vehicles in the section before the delay occurs. The historical passing speeds and the current section time-series speeds of the vehicles contain rich information. The historical passing speeds reflect the driving performances of the vehicles on different sections in the past. By analyzing the historical speed characteristics of the vehicles, the driving habits and speed change ranges of the vehicles can be understood.

[0099] Step 42: Construct the characteristics of the historical passing speed of vehicle i to obtain the historical passing speed v of vehicle i B , and the expression is as follows

[0100]

[0101] Among them, is the average passing speed of vehicle i in the immediately adjacent upstream section (the previous section), is the average passing speed of vehicle i in the upstream section before the previous one (the previous two sections).

[0102] Step 43: Construct the vehicle type characteristics, and the expression is as follows:

[0103] c i =(γ1, γ2, γ3) T (9)

[0104] Among them, γ1 represents the vehicle type, γ2 represents the vehicle axle speed, and γ3 represents the distance between the accident point and the previous ETC gantry

[0105] There are differences in the driving speeds and driving habits of different types of vehicles, and these differences will affect the time when the vehicles reach the disaster-damaged location. Constructing vehicle characteristics can comprehensively consider these factors. The vehicle type characteristics cover information such as vehicle type and number of axles. The driving speeds and handling performances of large trucks and small passenger cars are different, and their impacts on the traffic flow are also different. Incorporating these characteristics into the time estimation model can improve the accuracy of tracing the occurrence time of disaster-damaged events.

[0106] Step 44: Extract the traffic flow structure characteristics of this section, and select the traffic flow structure characteristics of this section within 5 minutes before and after the delayed vehicle enters. The traffic flow structure characteristics consist of the following three main parts:

[0107] d=(δ1, δ2, δ3) T (10)

[0108] Among them, δ1 represents the proportion of passenger cars in the traffic flow, δ2 represents the proportion of trucks in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

[0109] Specifically, the traffic flow structure has a significant impact on the driving time of vehicles within a section. By constructing section characteristics, these influencing factors can be incorporated into the vehicle arrival time estimation model. Different proportions of passenger cars, trucks, and special vehicles in the traffic flow structure characteristics will lead to differences in traffic flow characteristics, thereby affecting the overall vehicle speed. If the proportion of trucks in a section is high, since the driving speed of trucks is relatively slow, it will reduce the overall vehicle speed of this section and increase the vehicle passing time.

[0110] Furthermore, step 5 specifically includes the following steps:

[0111] Step 51, use a machine learning framework to construct a model for estimating the occurrence time of highway disaster and damage events; input vehicle characteristics, section characteristics, and historical transaction trajectories into the model for training and inference to predict the time ΔT for vehicle i to reach the accident location. i ,

[0112]

[0113]

[0114] S = {v A , d}

[0115]

[0116] where I is the vehicle characteristics of vehicle i, S is the section characteristics, and Tr i is the historical transaction trajectory of vehicle i.

[0117] Step 52, calculate the average time for the vehicles arriving at the accident location extracted during this period.

[0118]

[0119] where ΔT is the time for the disaster and damage vehicle to reach the accident location from the ETC gantry upstream of the journey, ΔT i is the time for vehicle i to reach the accident location from the ETC gantry upstream of the journey, and n is the total number of vehicles within the calculation range.

[0120] Then the estimated occurrence time of the disaster and damage event is

[0121] T = T i + ΔT (15)

[0122] where T is the estimated occurrence time of the disaster and damage event, and T i is the time when the earliest delayed traffic flow passes through the upstream ETC gantry.

[0123] The present invention adopts the above technical solutions and has the following technical advantages: 1. Precise positioning of the location and time of disaster-damaged events: Given the problems of large errors and inaccuracies in the traditional positioning of disaster-damaged events relying on manual reports and mileage markers, the present invention converts the highway mileage markers into longitude and latitude coordinates, establishes a mapping table, combines the key information at the event site, and uses the GIS system to achieve precise positioning of the event location. At the same time, based on the ETC transaction flow data and vehicle driving characteristics, a model is constructed to calculate the time of occurrence of the disaster-damaged event, solving the defect of large errors in the traditional time determination method, and providing accurate spatio-temporal information for subsequent rescue, liability determination and other work.

[0124] 2. Efficient integration and utilization of multi-source data: Aiming at the situation of scattered storage, inconsistent formats and insufficient utilization of existing highway data, the present invention first standardizes the ETC gantry data, toll station transaction data and service area data to ensure data consistency, and then removes outliers through data cleaning to integrate multi-source data. Deeply mine the information such as vehicle driving trajectories and speeds in the ETC data, and conduct comprehensive analysis in combination with other data to give full play to the value of the data and provide comprehensive and accurate data support for the analysis of disaster-damaged events.

[0125] 3. Construct a precise time traceability model and technical system: Considering the lack of current time traceability technology for disaster-damaged events and the problem that existing methods cannot accurately consider various influencing factors, the present invention extracts vehicle characteristics (speed, type) and section characteristics (physics, traffic flow structure), and uses LightGBM 4.3.0 to construct a vehicle arrival time estimation model, and combines real-time stream processing technology to predict the arrival time in real time. This model can comprehensively consider the influence of various factors on the vehicle driving time, achieve precise traceability of the time of occurrence of the disaster-damaged event, and provide a scientific basis for traffic management decisions.

[0126] 4. Optimize the information query and processing process: Since the existing information query process for disaster-damaged events is cumbersome and inefficient, the present invention designs a special data query process, uses the established mapping table to quickly match the mileage marker with the longitude and latitude coordinates, and improves the matching success rate through an incremental search strategy. At the same time, optimize the data processing process to quickly obtain the vehicle's historical passing records and section information, improve the information query and processing efficiency, and ensure the timeliness of emergency response.

[0127] Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Without conflict, the embodiments in the present application and the features in the embodiments may be combined with each other. Generally, the components of the embodiments of the present application described and illustrated in the accompanying drawings herein may be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application is not intended to limit the scope of the present application claimed, but merely represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the scope of protection of the present application.

Claims

1. A method for estimating the occurrence time of highway disaster events considering the traffic time lag effect, characterized by: It includes the following steps: Step 1: Collect key information from the incident site to form an emergency data sequence; transmit the collected emergency data sequence to a data center equipped with a GIS system for processing and analyzing the received data; Step 2: The data center matches the emergency event data sequence with the database, converts the mileage pile number in the emergency event data sequence into the latitude and longitude of the disaster event location, and accurately locates the event location; determines the upstream and downstream ETC gantry numbers, and calculates the distance γ3 between the event location and the upstream ETC gantry through the GIS system; Step 3: By comparing the vehicle traffic records of the downstream ETC gantry with the historical vehicle traffic data, a list of vehicles delayed due to the disaster is screened out; Step 4: Construct the vehicle feature vector set F of the disaster event section <f1,f2,…,f n >, where n is the total number of vehicles and the feature vector f of each vehicle i = i ; S; Tr i >, is the characteristic of vehicle i, specifically the historical speed of vehicle i and type characteristic c of vehicle i i ; S = <v A ,d> is the section feature, and the section feature S includes the historical speed feature v of the corresponding section A and the traffic flow structure characteristics d of the corresponding section;​ And the historical trajectory data Tr of vehicle i i , historical trajectory data Tr i Includes all ETC gantries that the vehicle passes through and transaction times; Step 5: Use the machine learning framework to build a model for estimating the time of occurrence of highway disaster events; i The feature vector of the historical transaction trajectory of vehicle i is used as the input to estimate the time when highway disasters occur. The model is trained and inferred to obtain the time ΔT when vehicle i arrives at the scene of the incident. i , and perform weighted average The estimated time of the disaster event is T = T i +ΔT is the estimated time of occurrence of the disaster event.

2. The method for estimating the occurrence time of highway disaster events considering the traffic time lag effect according to claim 1 is characterized by: The key information of the incident scene in step 1 includes the pile number information l, reporting time t, event type u; the emergency event data sequence x i = <l i ,t i ,u i >, where i is the number of the data sequence x, u i The event type information is used to identify the nature of the event.

3. The method for estimating the occurrence time of highway disaster events considering traffic time lag effect according to claim 1 is characterized by: In step 1, the data center stores the emergency event data sequence in the historical record database to support subsequent event analysis and emergency response.

4. The method for estimating the occurrence time of highway disaster events considering traffic time lag effect according to claim 1, characterized in that: Step 3 specifically includes the following steps: Step 31: By comparing the vehicle traffic records of the downstream ETC gantry with the historical vehicle traffic data, the vehicles delayed due to the disaster are screened out; Step 32: According to the vehicle license plate number p i Extract the historical ETC gantry transaction records of vehicle i from the ETC transaction flow data, and generate the ETC trajectory data of each trip; Step 33: Sort and aggregate by transaction time to form the ETC trajectory data of vehicle flow i in trip j in, The ETC gantry number that vehicle i passes through during trip j; The time it takes for vehicle i to pass through each ETC gantry; Step 34: after generating the ETC trajectory data, multiple trip data of vehicle i are integrated to form the complete historical trajectory data Tr of the corresponding vehicle. i , the specific expression is as follows: Where m is the total number of historical trips of vehicle i; Step 35, the historical trajectory data of all vehicles constitutes a data set eTr; eTr=<Tr1,Tr2,…,Tr n >(3) Where n is the total number of vehicles within the calculation range.

5. The method for estimating the occurrence time of highway disaster events considering the traffic time lag effect according to claim 4 is characterized by: In step 31, select vehicles whose entry time is no more than 1 minute different from the first batch of delayed vehicles to build a vehicle list where p i is the license plate number of vehicle i, is the time when vehicle i passes through the upstream ETC gantry.

6. The method for estimating the occurrence time of highway disaster events considering the traffic time lag effect according to claim 4 is characterized by: Step 4 specifically includes the following steps: Step 41: construct the segment historical speed feature v A , the expression is as follows: v A ＝(α1,α2,α3) 2 (4) α1=max(v1,v i ,…,v n )(5) α1=min(v1,v2,…,v n )(6) Among them, α1 is the maximum speed of the section, α2 is the minimum speed of the section, α3 is the average speed of the vehicles in the section, v1, v2, …, v n is the speed of each vehicle in the segment before the delay occurs; Step 42: construct features for the historical speed of vehicle i to obtain the historical speed v of the vehicle B , the expression is as follows in, is the average speed of vehicle i in the immediately upstream section, is the average speed of vehicle i in the previous upstream section; Step 43: Construct the vehicle type feature c i , the expression is as follows: c i =(γ1,γ2,γ3) T (9) Among them, γ1 represents the vehicle type, γ2 represents the vehicle axle speed, and γ3 represents the distance between the accident point and the previous ETC gantry; Step 44, obtaining the proportion of different types of vehicles in the same time slice in the same section to form a traffic flow structure feature, and using the corresponding traffic flow structure feature as the section feature; d=(δ1,δ2,δ3) T (10) Among them, δ1 represents the proportion of passenger cars in the traffic flow, δ2 represents the proportion of trucks in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

7. The method for estimating the occurrence time of highway disaster events considering traffic time lag effect according to claim 1 is characterized by: Step 5 specifically includes the following steps: Step 51, use a machine learning framework to build a model for estimating the time of occurrence of highway disaster events; input vehicle characteristics, section characteristics and historical transaction trajectories into the model, perform training and inference, and predict the time ΔT when vehicle i arrives at the scene of the incident i , ΔT i =F(I i ;S;Tr i )(11) I i ={v Bi ,c i } (12) S={v4,d} Among them, I is the vehicle characteristic of vehicle i, S is the segment characteristic, Tr i is the historical transaction track of vehicle i; Step 52, calculate the average time it takes for the vehicles extracted in the corresponding time period to arrive at the accident site, Among them, ΔT is the time it takes for the damaged vehicle to reach the accident site from the upstream ETC gantry, ΔT i is the time taken by vehicle i to reach the accident site from the upstream ETC gantry of the trip, and n is the total number of vehicles within the calculation range; Step 52, the estimated time of occurrence of the disaster event, specifically expressed as T=T i +ΔT(15) Where T is the estimated time of occurrence of the disaster event, T i The time when the earliest delayed traffic passes through the upstream ETC gantry.

8. The method for estimating the occurrence time of highway disaster events considering the traffic time lag effect according to claim 1 or 7, characterized in that: In step 5, the machine learning framework uses LightGBM3.4.0 to build a model for estimating the time when highway disaster events occur.

Images