Highway vehicle arrival time real-time estimation method based on ETC data

Abstract

This invention discloses a real-time estimation method for highway vehicle arrival times based on ETC data. By combining ETC transaction log data and vehicle GPS data, and matching ETC gantry transaction times with GPS data, corresponding GPS trajectory points can be quickly extracted, thus solving the tedious problem of matching a large number of trajectory points using traditional GPS data. The combination of GPS data enables accurate entry detection, overcoming the shortcomings of relying solely on traditional methods to identify whether a vehicle has entered a service area. This invention uses highway ETC transaction data to construct a highway gantry arrival time estimation model, and through real-time data, achieves real-time estimation of the arrival time of vehicles on the highway at the next gantry, meeting the monitoring and safety requirements of vehicles on the highway.

Description

Technical Field

[0001] This invention relates to the field of highway technology, and in particular to a method for real-time estimation of highway vehicle arrival time based on ETC data. Background Technology

[0002] With the large-scale deployment of ETC (Electronic Toll Collection) systems on highways in recent years, initial full coverage has been achieved. However, to further leverage the functions of the ETC system and create smart highways, it is necessary to conduct research on key technologies such as vehicle-road cooperation, utilize big data and artificial intelligence technologies to expand the functional applications of ETC, establish a vehicle-road perception and cooperation mechanism, and achieve refined supervision, operation, and service of highways. Therefore, it is urgent to achieve accurate control over vehicles and road conditions, significantly improving the decision-making efficiency of management departments. Furthermore, providing the public with real-time traffic information services and intelligent route planning through multiple channels will greatly optimize the public's highway travel experience.

[0003] With the continuous improvement of the expressway network, the expressway transportation industry is experiencing rapid development. Expressways have high traffic density and high speeds, resulting in traffic accidents with a wide impact and severe casualties. For expressway management units, understanding the driving status on expressways is fundamental to improving road safety, strengthening vehicle control capabilities, and facilitating emergency rescue. However, the driving status of vehicles on expressways is a "black box," leading to difficulties in traffic guidance, slow emergency response, and weak proactive control capabilities. With the continuous increase in the total mileage of expressways and the booming development of the expressway transportation industry, the management and monitoring of vehicles on expressways has become increasingly urgent. Vehicles traveling at high speeds in the complex road network not only pose potential risks to road safety but also significantly impact the efficiency and smoothness of the entire transportation network. Therefore, understanding the travel time of vehicles on expressways plays a crucial role in traffic planning, design, management, and travel information services. It provides data references for the implementation of traffic guidance schemes and the dissemination of travel information, serving as the most direct basis for travelers to make travel decisions.

[0004] To date, research methods for time-of-arrival (TOA) estimation primarily focus on utilizing GPS data, checkpoint data, and radar data, mainly falling into three categories: The first category involves collecting traffic flow parameters using coils and radar at locations and sections, typically with a coverage area of ​​around 150 meters; the second category uses floating car trajectory data to measure actual vehicle movement; and the third category uses checkpoints and other equipment to collect location vehicle characteristics. However, these methods primarily focus on TOA estimation within urban road networks, with insufficient research on highway ETC data. Specifically: 1. GPS and other positioning information are not yet accurate: Currently, vehicle positioning data mainly relies on Beidou vehicle-mounted terminals installed in vehicles, which send GPS positioning data to the platform at certain time frequencies for monitoring. On highways, GPS information loss or signal delays occur due to factors such as tunnels, mountains, and trees, making it more of a post-event processing method. 2. Traditional service area entry detection methods are relatively complex or have drawbacks: Since service areas may exist between highway gantries, vehicle entry into service areas may lead to inaccurate travel time estimations. Traditional entry detection methods rely solely on GPS data or service area snapshots. However, due to the short positioning intervals, each vehicle's GPS data generates a large number of trajectory points. Therefore, GPS-based entry detection requires extensive map matching operations. Using only service area snapshots for entry detection depends on entry probability or cameras before and after the service area, and varying weather conditions can lead to inaccuracies in license plate recognition. Traditional time-of-arrival (TOA) estimation methods perform poorly on highways: these methods are typically suitable for public transportation or specific sectors like cargo ships and airplanes, where routes and stops are usually well-defined, making it easy to estimate the entire journey time. However, the situation is more complex on highways, where routes are often not fixed, and highway gantries may include multiple stops and service areas. Therefore, traditional TOA methods are no longer applicable in this context, necessitating new approaches to meet the monitoring and safety requirements of vehicles en route on highways. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time estimation method for highway vehicle arrival time based on ETC data, so as to quickly detect whether a vehicle has entered a service area.

[0006] The technical solution adopted in this invention is:

[0007] A real-time estimation method for highway vehicle arrival times based on ETC data includes the following steps:

[0008] Step 1: Obtain ETC transaction data and GPS trajectory data, and extract GPS positioning data within the target segment as multi-source fusion data;

[0009] Step 2: Determine whether each trajectory of the target vehicle in the fused multi-source data has a stop state in the service area based on the time sequence; if so, classify the corresponding trajectory into the non-operational event set; otherwise, classify the corresponding trajectory into the operation event set.

[0010] Step 3: Extract vehicle features and segment features from the set of operational events. Vehicle features include speed features and type features. Use machine learning or deep learning algorithms to build a vehicle arrival time estimation model. Input the extracted vehicle features and segment features into the vehicle arrival time estimation model to estimate the arrival time.

[0011] Step 4: Obtain real-time ETC gantry transaction data, clean the data, and then distribute it using the vehicle's unique access identifier and license plate number;

[0012] Step 5: Use real-time stream processing technology to predict arrival times in real time using a vehicle arrival time estimation model. Specifically, relevant data will be transmitted through a high-throughput distributed publish-subscribe messaging system, and then real-time stream processing technology will be used to predict arrival times in real time.

[0013] Furthermore, step 1 specifically includes the following steps:

[0014] Step 1-1: Obtain ETC transaction data and GPS trajectory data, and perform data cleaning and transformation respectively;

[0015] Steps 1-2: Obtain ETC transaction data corresponding to license plate information from vehicle GPS data, and determine the transaction time of ETC gantry before and after the target section;

[0016] Steps 1-3 extract GPS positioning data from the GPS data within the time range during which the target vehicle interacts with the front and rear gantries of the target section.

[0017] Furthermore, in step 2, a STOP / MOVE model is constructed to describe the switching process of the trajectory between driving mode and stationary mode, and the homomorphic trajectory sequence is clustered according to the time sequence using STOP / MOVE state labels.

[0018] Furthermore, step 2 specifically includes the following steps:

[0019] Step 2-1: Obtain trajectory data from the fused multi-source data according to the time sequence;

[0020] Step 2-2: The STOP / MOVE model determines whether the vehicle speed on the current trajectory is 0; if so, proceed to step 2-5; otherwise, proceed to step 2-3.

[0021] Steps 2-3: Check if the current vehicle status and the label in the historical status list are in a stationary state; if so, obtain the start and end times from the status list and calculate the current travel time; otherwise, keep the current vehicle status label unchanged.

[0022] Steps 2-4: Determine if the movement time is greater than the dwell time threshold; if so, change the current vehicle status label to driving status and store it in the status list; otherwise, keep the current vehicle status label unchanged.

[0023] Step 2-5: Determine if the vehicle is within the service area; if so, proceed to step 2-6; otherwise, proceed to step 2-8.

[0024] Steps 2-6: Check if the current vehicle status label is in a driving state; if so, obtain the current speed start time and end time, and calculate the current dwell time; otherwise, keep the current vehicle status label unchanged.

[0025] Steps 2-7: Determine if the current dwell time is greater than the movement time threshold; if so, change the current vehicle status label to stationary; otherwise, keep the current vehicle status label unchanged.

[0026] Step 2-8: Determine whether all trajectories have been detected; if yes, proceed to step 2-9; otherwise, proceed to step 2-1.

[0027] Steps 2-9 involve correcting the vehicle's trajectory for different states and extracting the trajectory for each state, classifying it, and storing it in the corresponding event set.

[0028] Furthermore, the threshold for movement time is set to 120 seconds and the threshold for dwell time is set to 300 seconds.

[0029] Furthermore, in steps 2-9, the VSLC algorithm is used to correct the vehicle's trajectory under different states.

[0030] Furthermore, in steps 2-9, trajectories with the current vehicle status label set as "driving" are extracted and included in the operational event set, while trajectories with the current vehicle status label set as "stationary" are extracted and included in the non-operational event set.

[0031] Furthermore, step 3 specifically includes the following steps;

[0032] Step 3-1: Extract the vehicle's speed features. The specific steps are as follows:

[0033] Step 3-1-1: Construct features from the historical traffic speed and current segment temporal speed of vehicles, as shown in the following expression:

[0034]

[0035] in, These are the segment average traffic speed characteristics of the preceding time slices T1 and T2, respectively;

[0036] Step 3-1-2: Construct the vehicle's historical speed feature v A The expression is as follows:

[0037] v A =(α1,α2,α3,α4) T (2)

[0038] Where α1~α3 represent the historical driving speed characteristics of vehicles excluding service area sections, α1=max(v1,v2,...,v n ) represents the maximum speed of the vehicle in the section; α2=min(v1,v2,...,v n () indicates the minimum speed of a vehicle traveling in a given section; To represent the average driving speed of the vehicle to be estimated within a section; This represents the average speed of vehicles traveling through a section of highway.

[0039] Step 3-1-3: The temporal speed features of a segment provide important information about the current traffic conditions. By extracting the segment speed features, the impact of traffic conditions on travel time can be identified. The segment temporal features are constructed as follows:

[0040] v B =(β1,β2,β3,β4) T (3)

[0041] Wherein, β1~β4 represent the historical time slice velocity characteristics of the current segment.

[0042] Specifically, the day is divided into 288 time slices, each 5 minutes long. The segment speed of the current highway segment is calculated based on the transaction time when a vehicle enters the segment. The relationships between β1, β2, β3, and β4 are as follows: Figure 2 As shown in the time slice, Ts is defined. i Let β0 be the vehicle's current time slice, and β1 be the vehicle's edge velocity within this time slice. i-1 (V) represents the segment speed of the vehicle one time slice backward from the current segment, β2=Ts i-2(V) represents the segment speed of the vehicle two time slices prior to the current segment, β3=Ts i-3 (V) represents the segment speed of the vehicle three time slices prior to the current segment, β4=Ts i-4 (V) represents the segment speed of the vehicle four time slices prior to the current segment.

[0043] Step 3-2, construct the vehicle type features, the expression is as follows:

[0044] c = (γ1, γ2) T (4)

[0045] Where γ1 represents the vehicle type and γ2 represents the number of axles;

[0046] Step 3-3: Extract segment features. Segment features mainly include physical features and traffic flow structure features. Traffic flow structure features are the traffic flow structure features within the same time slice within the same segment.

[0047] The physical characteristics are expressed as follows:

[0048] D=(θ1) T (5)

[0049] Where θ1 represents the segment distance:

[0050] The expression for the traffic flow structure characteristics is as follows:

[0051] f = (δ1, δ2, δ3) T (6)

[0052] Among them, δ1 represents the proportion of passenger cars in the traffic flow, δ2 represents the proportion of freight cars in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

[0053] Specifically, traffic flow and traffic structure within a segment have a significant impact on the estimation of travel time for vehicles within that segment. Higher traffic flow increases congestion and delays, while the proportion and characteristics of different types of vehicles in the traffic flow structure also affect travel time variations. Traffic flow exhibits strong temporal and spatial characteristics; therefore, when calculating the traffic flow structure of a segment, the traffic flow within the same time slice within the same segment is considered as a characteristic of the vehicle arrival time estimation algorithm.

[0054] This invention employs the above technical solutions to quickly detect whether a vehicle has entered a service area. By combining ETC transaction data and vehicle GPS data, and matching ETC gantry transaction time with GPS data, this invention can quickly capture corresponding GPS trajectory points, thus solving the cumbersome problem of matching a large number of trajectory points using traditional GPS data. Furthermore, relying solely on service area camera data cannot accurately pinpoint whether a vehicle has entered the service area. Traditional methods typically rely on entry rates or camera license plate recognition to determine entry, but due to weather and installation angle factors, camera license plate recognition can be unreliable. Therefore, combining GPS data enables accurate entry detection, compensating for the shortcomings of traditional methods. This invention uses highway ETC transaction data to construct a model for estimating the arrival time of vehicles en route to gantry points on highways. Simultaneously, through real-time data, it achieves real-time estimation of the arrival time of vehicles en route to the next gantry on highways, meeting the monitoring and safety requirements of vehicles en route on highways.

[0055] This invention overcomes the cumbersome issues of traditional GPS data matching and addresses the shortcomings of relying solely on service area snapshot data for accurate positioning. Especially when service areas are equipped with cameras, it reduces the reliance on camera-based license plate recognition and the impact of varying weather conditions on license plate recognition accuracy. Utilizing modern technologies such as machine learning and deep learning, this invention successfully achieves real-time arrival time prediction for vehicles en route on highways, providing robust data support for monitoring and safety management of vehicles on highways. Attached Figure Description

[0056] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0057] Figure 1 This is a schematic diagram of the multi-source data fusion process of the present invention;

[0058] Figure 2 A flowchart illustrating the process for rapid entry inspection into the service area;

[0059] Figure 3 A schematic diagram illustrating the principle of service area entry inspection;

[0060] Figure 4 This is a schematic diagram of ETC trajectory data extraction.

[0061] Figure 5 This is a schematic diagram of ETC segment data extraction;

[0062] Figure 6 A schematic diagram of the velocity characteristics of historical time slices;

[0063] Figure 7 A schematic diagram illustrating the construction process of the arrival time estimation model;

[0064] Figure 8 This is a flowchart of the real-time prediction module. Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0066] Travel time is a crucial parameter in traffic flow characteristics, an important component of intelligent transportation systems, and one of the most important indicators for travelers. Estimating vehicle arrival times allows for the description of vehicle status during their journey. Accurate and reliable travel time estimates in transportation networks are essential for adjusting traffic flow according to time-related rules, helping travelers choose the best route before and even during their trip. Furthermore, the travel time index is a quantitative indicator of road segment congestion, enabling the study of highway segment congestion patterns under different levels of traffic accident characteristics and the determination of congestion severity. Because service areas exist between highway gantries, vehicles entering service areas may lead to inaccurate travel time estimates.

[0067] like Figures 1 to 8 As shown in the figure, this invention discloses a real-time estimation method for highway vehicle arrival time based on ETC data, including the following steps:

[0068] Step 1: Obtain ETC transaction data and GPS trajectory data, and extract GPS positioning data within the target segment as multi-source fusion data;

[0069] Step 2: Determine whether each trajectory of the target vehicle in the fused multi-source data has a stop state in the service area based on the time sequence; if so, classify the corresponding trajectory into the non-operational event set; otherwise, classify the corresponding trajectory into the operation event set.

[0070] Step 3: Extract vehicle features and segment features from the set of operational events. Vehicle features include speed features and type features. Use machine learning or deep learning algorithms to build a vehicle arrival time estimation model. Input the extracted vehicle features and segment features into the vehicle arrival time estimation model to estimate the arrival time.

[0071] Step 4: Obtain real-time ETC gantry transaction data, clean the data, and then distribute it using the vehicle's unique access identifier and license plate number;

[0072] Step 5: Use real-time stream processing technology to predict arrival times in real time using a vehicle arrival time estimation model. Specifically, relevant data will be transmitted through a high-throughput distributed publish-subscribe messaging system, and then real-time stream processing technology will be used to predict arrival times in real time.

[0073] Furthermore, step 1 specifically includes the following steps:

[0074] Step 1-1: Obtain ETC transaction data and GPS trajectory data, and perform data cleaning and transformation respectively;

[0075] Steps 1-2: Obtain ETC transaction data corresponding to license plate information from vehicle GPS data, and determine the transaction time of ETC gantry before and after the target section;

[0076] Steps 1-3 extract GPS positioning data from the GPS data within the time range during which the target vehicle interacts with the front and rear gantries of the target section.

[0077] Furthermore, in step 2, a STOP / MOVE model is constructed to describe the switching process of the trajectory between driving mode and stationary mode, and the homomorphic trajectory sequence is clustered according to the time sequence using STOP / MOVE state labels.

[0078] Furthermore, step 2 specifically includes the following steps:

[0079] Step 2-1: Obtain trajectory data from the fused multi-source data according to the time sequence;

[0080] Step 2-2: The STOP / MOVE model determines whether the vehicle speed on the current trajectory is 0; if so, proceed to step 2-5; otherwise, proceed to step 2-3.

[0081] Steps 2-3: Check if the current vehicle status and the label in the historical status list are in a stationary state; if so, obtain the start and end times from the status list and calculate the current travel time; otherwise, keep the current vehicle status label unchanged.

[0082] Steps 2-4: Determine if the movement time is greater than the dwell time threshold; if so, change the current vehicle status label to driving status and store it in the status list; otherwise, keep the current vehicle status label unchanged.

[0083] Step 2-5: Determine if the vehicle is within the service area; if so, proceed to step 2-6; otherwise, proceed to step 2-8.

[0084] Steps 2-6: Check if the current vehicle status label is in a driving state; if so, obtain the current speed start time and end time, and calculate the current dwell time; otherwise, keep the current vehicle status label unchanged.

[0085] Steps 2-7: Determine if the current dwell time is greater than the movement time threshold; if so, change the current vehicle status label to stationary; otherwise, keep the current vehicle status label unchanged.

[0086] Step 2-8: Determine whether all trajectories have been detected; if yes, proceed to step 2-9; otherwise, proceed to step 2-1.

[0087] Steps 2-9 involve correcting the vehicle's trajectory for different states and extracting the trajectory for each state, classifying it, and storing it in the corresponding event set.

[0088] Furthermore, the threshold for movement time is set to 120 seconds and the threshold for dwell time is set to 300 seconds.

[0089] Furthermore, in steps 2-9, the VSLC algorithm is used to correct the vehicle's trajectory under different states.

[0090] Furthermore, in steps 2-9, trajectories with the current vehicle status label set as "driving" are extracted and included in the operational event set, while trajectories with the current vehicle status label set as "stationary" are extracted and included in the non-operational event set.

[0091] Furthermore, step 3 specifically includes the following steps;

[0092] Step 3-1: Extract the vehicle's speed features. The specific steps are as follows:

[0093] Step 3-1-1: Construct features from the historical traffic speed and current segment temporal speed of vehicles, as shown in the following expression:

[0094]

[0095] in, These are the segment average traffic speed characteristics of the preceding time slices T1 and T2, respectively;

[0096] Step 3-1-2: Construct the vehicle's historical speed feature v A The expression is as follows:

[0097] v A =(α1,α2,α3,α4) T (2)

[0098] Where α1~α3 represent the historical driving speed characteristics of vehicles excluding service area sections, α1=max(v1,v2,...,v n ) represents the maximum speed of the vehicle in the section; α2=min(v1,v2,...,v n () indicates the minimum speed of a vehicle traveling in a given section; To represent the average driving speed of the vehicle to be estimated within a section; This represents the average speed of vehicles traveling through a section of highway.

[0099] Step 3-1-3: The temporal speed features of a segment provide important information about the current traffic conditions. By extracting the segment speed features, the impact of traffic conditions on travel time can be identified. The segment temporal features are constructed as follows:

[0100] v B =(β1,β2,β3,β4) T (3)

[0101] Wherein, β1~β4 represent the historical time slice velocity characteristics of the current segment.

[0102] Specifically, such as Figure 2 As shown on the timeline, a day is divided into 288 time slices with a length of 5 minutes. The segment speed of the current highway segment is calculated based on the transaction time when a vehicle enters the segment. The relationship between β1, β2, β3, and β4 is as follows: Figure 2 As shown in the time slice, Ts is defined. i Let β0 be the vehicle's current time slice, and β1 be the vehicle's edge velocity within this time slice. i-1 (V) represents the segment speed of the vehicle one time slice backward from the current segment, β2=Ts i-2 (V) represents the segment speed of the vehicle two time slices prior to the current segment, β3=Ts i-3 (V) represents the segment speed of the vehicle three time slices prior to the current segment, β4=Ts i-4 (V) represents the segment speed of the vehicle four time slices prior to the current segment.

[0103] Step 3-2, construct the vehicle type features, the expression is as follows:

[0104] c = (γ1, γ2) T (4)

[0105] Where γ1 represents the vehicle type and γ2 represents the number of axles;

[0106] Step 3-3: Extract segment features. Segment features mainly include physical features and traffic flow structure features. Traffic flow structure features are the traffic flow structure features within the same time slice within the same segment.

[0107] The physical characteristics are expressed as follows:

[0108] D=(θ1) T (5)

[0109] Where θ1 represents the segment distance:

[0110] The expression for the traffic flow structure characteristics is as follows:

[0111] f = (δ1, δ2, δ3) T (6)

[0112] Among them, δ1 represents the proportion of passenger cars in the traffic flow, δ2 represents the proportion of freight cars in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

[0113] Specifically, traffic flow and traffic structure within a segment have a significant impact on the estimation of travel time for vehicles within that segment. Higher traffic flow increases congestion and delays, while the proportion and characteristics of different types of vehicles in the traffic flow structure also affect travel time variations. Traffic flow exhibits strong temporal and spatial characteristics; therefore, when calculating the traffic flow structure of a segment, the traffic flow within the same time slice within the same segment is considered as a characteristic of the vehicle arrival time estimation algorithm.

[0114] The specific principles of this invention will be explained in detail below:

[0115] like Figure 1 The multi-source data fusion algorithm flow is shown below. First, it acquires ETC transaction data corresponding to the license plate information from the vehicle's GPS data. Second, since GPS data cannot be directly mapped one-to-one with ETC segments, this algorithm employs a constraint based on the transaction time of the ETC gantries before and after the segment. Finally, through this constraint, GPS positioning data that matches the target vehicle's information interaction with the gantries before and after the segment can be further extracted from the GPS data, thus achieving the acquisition of GPS positioning data within the target segment.

[0116] like Figure 2 As shown in Figure 3, the service area rapid entry detection algorithm flow is as follows. Vehicle driving behavior can be categorized into two modes based on its motion state: driving mode and stationary mode. The trajectory in stationary mode has multiple semantic meanings, such as stops during traffic congestion, refueling, dining, or rest. Considering subsequent travel time, this invention divides the trajectory into two main categories: driving state (referring to normal driving on the highway) and non-driving state (including rest, dining, etc.).

[0117] This invention utilizes a STOP / MOVE model to describe the switching process of a trajectory between these two states. The diagram illustrates the running state of a trajectory. This invention marks trajectories with a speed greater than 0 as the Move state and trajectories with a speed of 0 as the Stop state. As can be seen from the diagram, when a vehicle passes through a service area, its state switches from moving to stationary.

[0118] Based on the above, this invention proposes a fast service area entry detection algorithm. First, this invention sets up a service area buffer to facilitate the determination of whether entry into the service area has occurred. Based on time sequence, this invention uses STOP / MOVE status labels to cluster homomorphic trajectory sequences. Here, this invention sets a movement duration threshold of 120 seconds and a dwell time threshold of 300 seconds. Using these two thresholds, this invention switches the labels for operational or non-operational states with durations shorter than these thresholds. This invention employs the VSLC algorithm to correct the trajectory status and extracts and stores these trajectories into operational event sets and non-operational event sets respectively.

[0119] This invention proposes a model for real-time estimation of the arrival time of vehicles en route to highway gantries, specifically including a data processing module, a service area rapid entry detection module, an arrival time estimation model construction module, and a real-time prediction module.

[0120] (1) The data processing module includes: ETC transaction data cleaning, GPS trajectory data cleaning, ETC trajectory data generation, and ETC segment data generation.

[0121] 1.1) ETC transaction data cleaning: ETC transaction data anomalies mainly include errcode anomalies, vehicle identification anomalies, and vehicle type anomalies. The data filtering method used in this invention addresses these three types of anomalies.

[0122] 1.2) GPS Trajectory Data Cleaning: Due to the influence of satellite geometry, receiver errors, and noise, some GPS data points drift outside the highway network. This drift affects the regularity of vehicle position changes and the model's fitting effect. Therefore, for GPS data points with repeated positioning of the same vehicle, excessively long positioning time intervals, continuously changing speeds but unchanged vehicle positions, continuously 0 km / h speeds with unchanged vehicle positions, and drifting trajectories outside the highway, the MapProvider API is used for drift point cleaning.

[0123] 1.3) ETC trajectory data generation: such as Figure 4 As shown, since ETC transaction data contains unique trajectory identifiers, ETC trajectory data can be obtained by sorting and aggregating transaction times.

[0124] 1.4) ETC segment data generation: such as Figure 5 As shown, after the ETC trajectory is generated, the gantry transaction information and gantry transaction time in the trajectory data can be recombined in pairs to obtain the ETC segment data.

[0125] (2) The service area fast entry detection module includes: multi-source data fusion algorithm and service area fast entry detection algorithm.

[0126] (3) Figure 7 As shown, the arrival time estimation model building module includes: vehicle feature and segment feature extraction, and model building.

[0127] 3.1) Vehicle and segment feature extraction, including the following:

[0128] 3.1-1) Speed ​​Characteristics: Speed ​​characteristics are one of the key features for estimating the travel time of high-speed vehicles. Vehicle speed is closely related to travel time; higher speeds generally mean shorter travel times, while lower speeds may lead to longer travel times. To estimate the arrival time of a vehicle, it is necessary to construct features based on the vehicle's historical travel speed and the current segment's time-series speed, as shown in formula (1):

[0129]

[0130] in These are the average passage speeds of the current segment in the preceding time slices T1 and T2, respectively.

[0131] Historical vehicle speed is an important feature for estimating vehicle arrival time. Constructing vehicle speed features v A As shown in formula (2).

[0132] v A =(α1,α2,α3,α4) T (2)

[0133] Among them, attributes α1 to α3 represent the historical driving speed characteristics of vehicles excluding service area sections, α1 = max(v1, v2, ..., v n ) represents the maximum speed of the vehicle in the section; α2=min(v1,v2,...,v n () indicates the minimum speed of a vehicle traveling in a given section; To represent the average driving speed of the vehicle to be estimated within a section; This represents the average speed of vehicles traveling on a section of highway.

[0134] The temporal speed characteristics of a traffic segment can provide important information about the current traffic conditions. By extracting the speed characteristics of the segment, this invention can identify the impact of traffic conditions on travel time. The temporal characteristics of the segment are constructed as shown in formula (3).

[0135] v B =(β1,β2,β3,β4) T (3)

[0136] Where β1~β4 represent the historical time slice velocity characteristics of the current segment, such as Figure 6As shown on the timeline, a day is divided into 288 time slices with a length of 5 minutes. The segment speed of the current highway segment is calculated based on the transaction time when a vehicle enters the segment. The relationship between β1, β2, β3, and β4 is as follows: Figure 6 As shown in the time slice, Ts is defined. i Let β0 be the vehicle's current time slice, and β1 be the vehicle's edge velocity within this time slice. i-1 (V) represents the segment speed of the vehicle one time slice backward from the current segment, β2=Ts i-2 (V) represents the segment speed of the vehicle two time slices prior to the current segment, β3=Ts i-3 (V) represents the segment speed of the vehicle three time slices prior to the current segment, β4=Ts i-4 (V) represents the segment speed of the vehicle four time slices prior to the current segment.

[0137] 3.1-2) Vehicle Characteristics: Considering the different travel behaviors of users with different vehicle types, vehicle type is also an important characteristic in vehicle arrival time estimation. Different vehicle types have a significant impact on vehicle arrival time prediction. For example, sedans typically have higher acceleration performance and lower air resistance, and may travel at higher speeds on highways. Conversely, trucks or heavy-duty vehicles, due to their larger size and weight, may require longer start-up and travel times. Therefore, the characteristics and performance of different vehicle types must be considered when predicting vehicle arrival times to improve the accuracy of the prediction. Their vehicle characteristics are shown in Formula (4).

[0138] c = (γ1, γ2) T (4)

[0139] Where γ1 represents the vehicle type and γ2 represents the number of axles.

[0140] 3.2) Segment characteristics: Arrival time estimation is inseparable from the extraction and analysis of segment characteristics of the highway segment through which the vehicle is currently traveling. Segment characteristics mainly include physical characteristics and traffic flow structure characteristics. Physical characteristics are shown in formula (5).

[0141] D=(θ1) T (5)

[0142] Where θ1 represents the segment distance.

[0143] Traffic flow and traffic structure within a segment have a significant impact on the estimation of travel time for vehicles within that segment. Higher traffic flow increases congestion and delays, while the proportion and characteristics of different types of vehicles in the traffic flow structure also affect the variation in travel time. Traffic flow has strong temporal and spatial characteristics; therefore, when calculating the traffic flow structure of a segment, the traffic flow within the same time slice within the same segment is considered as a characteristic of the vehicle arrival time estimation algorithm, as shown in Equation (6).

[0144] f = (δ1, δ2, δ3) T (6)

[0145] Where δ1 represents the proportion of passenger vehicles in the traffic flow, δ2 represents the proportion of freight vehicles in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

[0146] (4) The real-time prediction module includes: real-time data transfer and prediction result generation. For example... Figure 8 As shown, the three types of abnormal data mentioned above are removed during the transmission of the collected ETC gantry transaction data to the platform. The cleaned and transformed data is then distributed using the vehicle's unique access identifier and license plate number. This data is transmitted through a high-throughput distributed publish-subscribe messaging system, and then real-time stream processing technology is used to predict real-time arrival times.

[0147] This invention employs the above technical solutions to quickly detect whether a vehicle has entered a service area. By combining ETC transaction data and vehicle GPS data, and matching ETC gantry transaction time with GPS data, this invention can quickly capture corresponding GPS trajectory points, thus solving the cumbersome problem of matching a large number of trajectory points using traditional GPS data. Furthermore, relying solely on service area camera data cannot accurately pinpoint whether a vehicle has entered the service area. Traditional methods typically rely on entry rates or camera license plate recognition to determine entry, but due to weather and installation angle factors, camera license plate recognition can be unreliable. Therefore, combining GPS data enables accurate entry detection, compensating for the shortcomings of traditional methods. This invention uses highway ETC transaction data to construct a model for estimating the arrival time of vehicles en route to gantry points on highways. Simultaneously, through real-time data, it achieves real-time estimation of the arrival time of vehicles en route to the next gantry on highways, meeting the monitoring and safety requirements of vehicles en route on highways.

[0148] Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Without conflict, the embodiments and features in the embodiments of this application can be combined with each other. The components of the embodiments of this application described and illustrated herein can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

Claims

1. A real-time estimation method for highway vehicle arrival time based on ETC data, characterized in that: It includes the following steps: Step 1: Obtain ETC transaction data and GPS trajectory data, and extract GPS positioning data within the target segment as multi-source fusion data; Step 2: Determine whether each segment of the target vehicle's trajectory in the fused multi-source data has a stop status when entering a service area based on the time sequence; If so, the corresponding trajectory is assigned to the non-operational event set; otherwise, the corresponding trajectory is assigned to the operational event set. In step 2, a STOP / MOVE model is constructed to describe the switching process of the trajectory between driving mode and stationary mode. The homomorphic trajectory sequence is clustered according to the time sequence using STOP / MOVE state labels. Step 2 specifically includes the following steps: Step 2-1, obtain trajectory data from the fused multi-source data according to the time sequence. Step 2-2: The STOP / MOVE model determines whether the vehicle speed on the current trajectory is 0; if so, proceed to step 2-5; otherwise, proceed to step 2-3. Steps 2-3: Check if the current vehicle status and the label in the historical status list are in a stationary state; if so, obtain the start and end times from the status list and calculate the current travel time; otherwise, keep the current vehicle status label unchanged. Steps 2-4: Determine if the movement time is greater than the dwell time threshold; if so, change the current vehicle status label to driving status and store it in the status list; otherwise, keep the current vehicle status label unchanged. Step 2-5: Determine if the vehicle is within the service area; if so, proceed to step 2-6; otherwise, proceed to step 2-8. Steps 2-6: Check if the current vehicle status label and the label in the historical status list are in a driving state; if so, obtain the start time and end time in the status list and calculate the current dwell time; otherwise, keep the current vehicle status label unchanged. Steps 2-7: Determine if the current dwell time is greater than the movement time threshold; if so, change the current vehicle status label to stationary and store it in the status list; otherwise, keep the current vehicle status label unchanged. Step 2-8: Determine whether all trajectories have been detected; if so, proceed to step 2-9; otherwise, proceed to step 2-1; Step 2-9: Correct the trajectories of the vehicle in different states, and extract the trajectories of the corresponding states and store them in the corresponding event sets. Step 3: Extract vehicle features and segment features from the set of operational events. Vehicle features include speed features and type features, while segment features include physical features and traffic flow structure features. Use machine learning or deep learning algorithms to build a vehicle arrival time estimation model. Input the extracted vehicle features and segment features into the vehicle arrival time estimation model to estimate the arrival time. Step 4: Obtain real-time collected ETC transaction data, clean the data, and then distribute it using the vehicle's unique access identifier and license plate number; Step 5: Use real-time stream processing technology to predict arrival time in real time using a vehicle arrival time estimation model.

2. The real-time estimation method for highway vehicle arrival time based on ETC data according to claim 1, characterized in that: Step 1 specifically includes the following steps: Step 1-1: Obtain ETC transaction data and GPS trajectory data, and perform data cleaning and transformation respectively; Steps 1-2: Obtain ETC transaction data corresponding to license plate information from vehicle GPS data, and determine the transaction time of ETC gantry before and after the target section; Steps 1-3 extract GPS positioning data from the GPS data within the time range during which the target vehicle interacts with the front and rear gantries of the target section.

3. The real-time estimation method for highway vehicle arrival time based on ETC data according to claim 1, characterized in that: Set the movement time threshold to 120 seconds and the dwell time threshold to 300 seconds.

4. The real-time estimation method for highway vehicle arrival time based on ETC data according to claim 1, characterized in that: In steps 2-9, the VSLC algorithm is used to correct the vehicle's trajectory under different states.

5. The real-time estimation method for highway vehicle arrival time based on ETC data according to claim 1, characterized in that: In steps 2-9, trajectories with the current vehicle status label set to "driving" are extracted and included in the operational event set, while trajectories with the current vehicle status label set to "stationary" are extracted and included in the non-operational event set.

6. The real-time estimation method for highway vehicle arrival time based on ETC data according to claim 1, characterized in that: Step 3 specifically includes the following steps; Step 3-1: Extract the vehicle's speed features. The specific steps are as follows: Step 3-1-1: Construct the speed feature v by analyzing the historical traffic speed and the current segment's time-series speed of the vehicle. The expression is as follows: in, These are the segment average traffic speed characteristics of the preceding time slices T1 and T2, respectively; Step 3-1-2: Construct the vehicle's historical speed feature v A The expression is as follows: v A ＝(α1,α2,α3,α4) T (2) Where α1~α3 represent the historical driving speed characteristics of vehicles excluding service area sections, α1=max(v1,v2,...,v n ) represents the maximum speed of the vehicle in the section; α2=min(v1,v2,...,v n () indicates the minimum speed of a vehicle traveling in a given section; To represent the average driving speed of the vehicle to be estimated within a section; This represents the average speed of vehicles traveling through a section of highway. Step 3-1-3: By extracting the speed features of the segments, the impact of traffic conditions on travel time is identified, and the expression for the segment time-series features is constructed as follows: v B ＝(β1,β2,β3,β4) T (3) Wherein, β1~β4 represent the historical time slice velocity characteristics of the current segment; Step 3-2, construct the vehicle type features, the expression is as follows: c=(γ1,γ2) T (4) Where γ1 represents the vehicle type and γ2 represents the number of axles; Step 3-3: Extract segment features. The traffic flow structure features are the traffic flow structure features within the same time slice of the same segment. The physical characteristics are expressed as follows: D=(θ1) T (5) Where θ1 represents the segment distance: The expression for the traffic flow structure characteristics is as follows: f=(δ1,δ2,δ3) T (6) Among them, δ1 represents the proportion of passenger vehicles in the traffic flow, δ2 represents the proportion of freight vehicles in the traffic flow, and δ3 represents the proportion of special vehicles in the traffic flow.

Images