A fine-grained traffic state estimation method based on ETC gantry data

Abstract

The present invention relates to a fine-grained traffic state estimation method based on ETC gantry data, including: dividing a road network into multiple scene units, dividing fine-grained vehicle types, selecting vehicle travel time and traffic flow as traffic state variables; screening and processing ETC gantry charging data, constructing a fine-grained vehicle set passing through each scene unit within a fixed time; calculating the average travel speed of vehicles of the same type passing through each scene unit; dividing dynamic time intervals, calculating ramp diversion coefficients; estimating the average travel time of vehicles of the same type passing through each scene unit in different dynamic time intervals; calculating the travel time of vehicles of each type to any target section downstream; predicting the future fine-grained traffic flow of the ETC gantry; aggregating fine-grained traffic flows in any target section downstream and at any time, and obtaining a traffic flow prediction value of a detection blind area. Compared with the prior art, the present invention has the advantages of high estimation accuracy, fine granularity and good timeliness.

Description

Technical Field

[0001] The present invention relates to the field of highway traffic information estimation, and in particular to a fine-grained traffic state estimation method based on ETC gantry data. Background Art

[0002] Traffic state estimation (TSE) is the process of estimating traffic variables on a road section based on prior traffic knowledge and partial observation data. It plays an important role in traffic tasks such as congestion control, ramp metering, risk detection, and accident emergency response. It is also a basic link in the construction of intelligent monitoring and management of highway networks.

[0003] Considering the differences in the speeds of different vehicles on the highway, existing studies have proposed to classify vehicles into large and small (or large, medium, and small) for estimation schemes. However, few studies have conducted fine-grained classification estimation for passenger and freight vehicles, which is often required in actual traffic management tasks, such as mixed traffic research of passenger and freight vehicles in detection blind spots and risk conflict detection. Secondly, exploring the spatial propagation effect of traffic flow is a key aspect to improve estimation accuracy. Existing studies use traditional detection data such as toll stations and circular coils to only emphasize the spatiotemporal modeling between upstream and downstream sensors on highways, but lack in-depth analysis of how factors such as ramp diversion and tunnel speed limits interact with traffic flow and travel time. Summary of the invention

[0004] The purpose of the present invention is to provide a fine-grained traffic state estimation method based on ETC gantry data, consider the fine-grained vehicle type classification, explore the spatial propagation effect of traffic flow, realize fine-grained traffic state estimation based on existing ETC data, and improve the estimation accuracy.

[0005] The purpose of the present invention can be achieved by the following technical solutions:

[0006] A fine-grained traffic state estimation method based on ETC gantry data includes the following steps:

[0007] Step 1) Divide the road network into multiple scene units, divide the fine-grained vehicle types, and select vehicle travel time and traffic flow as traffic state variables;

[0008] Step 2) Filter and process ETC gantry charging data to construct a fine-grained vehicle set that passes through each scene unit within a fixed time;

[0009] Step 3) calculating the travel time and average travel speed of individual vehicles passing through the scene units, and based on the results, calculating the average travel speed of the same type of vehicles passing through each scene unit;

[0010] Step 4) dividing the dynamic time interval and calculating the ramp diversion coefficient;

[0011] Step 5) estimating the average travel time of the same type of vehicles passing through each scene unit in different dynamic time intervals;

[0012] Step 6) calculating the travel time for each type of vehicle to reach any target section downstream based on the average travel time obtained in step 5);

[0013] Step 7) Select a spatiotemporal prediction model to predict the fine-grained traffic flow of the ETC gantry in the future time period;

[0014] Step 8) Based on the ramp diversion coefficient, the travel time of each type of vehicle to reach any target section downstream, and the predicted fine-grained traffic flow of the ETC gantry in the future time period, the fine-grained traffic flow of any target section downstream and at any time is aggregated to obtain the traffic flow prediction value of the detection blind spot.

[0015] In the step 1), the segmentation rule for segmenting the road network into multiple scene units is: taking the complex road section locations that may affect the evolution of traffic flow as specific segmentation nodes, and connecting any two adjacent segmentation nodes as scene units, wherein the complex road section locations include ETC gantries, toll stations, ramps, tunnel entrances and exits, bridges, and ramps.

[0016] The step 2) is specifically as follows:

[0017] Construct a set of k-type vehicles that pass through scene unit i within a fixed time (day, week, month) in, n i,k is the statistic of k-type vehicles passing through scene unit i, veh represents the vehicle, and β represents the scene type.

[0018] The step 3) is specifically as follows:

[0019] Calculate the travel time t(j) and average travel speed of individual vehicle j through scene unit i Based on this result, calculate the average speed of vehicles of the same type passing through each scene unit.

[0020]

[0021]

[0022] Among them, T O (j) is the recorded time when the vehicle arrives at the upstream node; T E (j) is the recorded time when the vehicle arrives at the downstream node; L is the distance traveled by the vehicle in scene unit i; n r,kis the statistics of k-type vehicles passing through the ramp; is the statistic of passing the ramp; n A,k is the statistics of k-type vehicles passing through the upstream ECT gantry, is the set of all k-type vehicles passing through the upstream ECT gantry obtained by statistics within a fixed time, up_Seg ub (i) down_Seg ub (i) Respectively represent the upstream and downstream sections without ETC.

[0023] In step 4), taking into account the periodicity of traffic flow and the significant differences in traffic conditions in different time periods, one day is dynamically divided into p time intervals, and based on the matching of license plate information of each ETC gantry, the source and destination of the vehicle are analyzed, and the vehicle driving route is restored to accurately analyze the spatial propagation effect of the traffic flow.

[0024] In step 4), the ramp diversion coefficient is calculated by: based on the different scene units divided in step 1), the exit ramp diversion coefficients R of different types of vehicles are calculated. off-ramp , entrance ramp diversion coefficient R on-ramp and the combined entrance and exit ramp split coefficient R ramp ,in,

[0025]

[0026]

[0027]

[0028] Among them, ρ t Indicates the time interval, N A,k (ρ t ) and N B,k (ρ t ) is the statistical time interval ρ t The collection of K-type vehicles passing through the upstream and downstream ETC gantries.

[0029] The step 5) is specifically as follows: based on the dynamic time interval divided in step 4), the average travel time of the k-type vehicle passing through the scene unit i is estimated

[0030]

[0031] in, represents the k-type car on scene unit i t The average driving speed of the time period is calculated according to different scene types; L i is the distance traveled by the vehicle in scene unit i.

[0032] The step 6) is specifically as follows: taking any ETC gantry as the timing starting point, based on the average travel time of step 5) According to the scene types of different scene units, the travel time of the k-type vehicle to reach any downstream target section D is calculated.

[0033] In the step 7), the prediction model may select the currently popular deep learning model based on graph convolutional network (GCN).

[0034] The step 8) is specifically as follows:

[0035] Based on the travel time of a k-type vehicle to reach any downstream target section D make And determine the required time T demand , the traffic flow prediction value of the detection blind area is determined according to two conditions:

[0036] i) When T demand ≤T traveling When the flow of k-type vehicles at the target location within the ΔT time period is determined based on historical statistical values

[0037] ii) When T demand >T traveling When the fine-grained traffic flow of the ETC gantry in the future time period predicted in step 7) is used to determine the flow of k-type vehicles at the target position within the ΔT time period

[0038] The method further comprises:

[0039] Step 9) Design a verification scheme for detecting vehicle travel time and traffic flow in the blind spot, and evaluate the estimation performance of each variable in each verification scheme. The verification scheme equates the ETC gantry to the target position, uses other ETC gantries to infer the traffic status of the equivalent position, and evaluates the corresponding model estimation performance.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] 1. The present invention uses ETC gantry equipment as source data, which has obvious advantages over traditional traffic data collection and investigation methods: first, ETC gantry data is collected automatically without manual intervention, saving manpower and time; second, data is collected in real time to provide accurate and timely traffic information, which is updated more quickly than traditional methods; third, it is recorded by an electronic system with high accuracy, reducing human errors; fourth, it has the advantage of big data analysis and high vehicle coverage.

[0042] 2. The present invention takes into account the different characteristics and driving behaviors of buses and trucks, and performs fine-grained classification estimation, providing refined data support for tasks such as mixed traffic research of buses and trucks and risk conflict detection in actual traffic management.

[0043] 3. Based on the influence of various factors such as ramp diversion, road line shape, tunnel speed limit on traffic flow estimation, the present invention proposes a systematic segmentation rule, divides the entire road network into different scene units, and calculates the diversion coefficient of each ramp, which can provide a deeper understanding of the spatial propagation effect of traffic flow. Embedding this process in the design algorithm can effectively improve the accuracy of traffic state estimation.

[0044] 4. The present invention proposes a fine-grained traffic state estimation method based on ETC gantry data, which solves the fine-grained and timeliness problems in highway traffic state estimation, and explores the impact of spatial propagation effects on traffic state estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 is a flow chart of the method of the present invention;

[0046] Figure 2 5 typical scene unit diagrams obtained based on segmentation rules in one embodiment;

[0047] Figure 3 A network framework diagram of a prediction model in an embodiment;

[0048] Figure 4 A schematic diagram of a fine-grained vehicle flow estimation procedure in an embodiment;

[0049] Figure 5 A design diagram of a verification scheme for a model in an embodiment. DETAILED DESCRIPTION

[0050] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0051] This embodiment provides a fine-grained traffic state estimation method based on ETC gantry data, such as Figure 1 As shown, the following steps are included:

[0052] Step 1) Divide the road network into multiple scene units and classify fine-grained vehicle types, and select vehicle travel time and traffic flow as traffic state variables.

[0053] Specifically, complex road sections that may affect the evolution of traffic flow, such as ETC gantries, toll stations, ramps, tunnel entrances and exits, bridges, and ramps, are taken as specific segmentation nodes, and any two adjacent segmentation nodes are connected as scene units.

[0054] The research objects selected in this example are distributed in 6 ETC gantries, 1 hub interchange, 3 ramp toll stations and 2 tunnel scenes. Based on this, combined with the distribution characteristics and flow transfer rules of actual highway ramps, the research objects are divided into five different types of scene units, such as Figure 2 shown.

[0055] Taking into account the differences in transportation requirements, vehicle characteristics and driving speed, vehicles are divided into two types: passenger cars and trucks. Based on the number of seats and vehicle length, passenger cars are further divided into large passenger cars. B1 、Medium-sized buses B2 、Small passenger car B3 Based on the length and total weight of the vehicle and cargo, passenger vehicles are further divided into large trucks. T1 , Medium Truck T2 , Minivan T3 .

[0056] Step 2) Filter and process the ETC gantry charging data to construct a fine-grained set of vehicles passing through each scene unit within a fixed time.

[0057] Step 21) Filter out the field information required for ETC gantry charging data, classify it into vehicle attributes, location attributes and time attributes, perform statistics in the order of collection time, complete abnormal data processing (assign Null to erroneous, repeated and fluctuating data, and then complete missing data), and perform fine-grained vehicle clustering at a time interval of 5 minutes.

[0058] Step 22) Based on the five scene units segmented in step 1, count the set of k-type vehicles passing through any scene unit i within a fixed time (day, week, month) in, n i,k is the statistic of k-type vehicles passing through scene unit i, is the set of vehicles passing through the ramp, and β represents the scene type. The specific calculation is as follows:

[0059]

[0060]

[0061] Among them, β is Figure 2 The subset of the five scenes Scene{1,2,3,4,5} shown, N A,k The number of nodes passing through the upstream Segment obtained by statistics within a fixed timeETC The set of all k-type vehicles, N B,k is the set of k-type vehicles passing through the downstream ETC gantry within a fixed time, up_Seg ub (i) down_Seg ub (i) Respectively represent the upstream and downstream sections without ETC.

[0062] Step 3) Calculate the travel time and average travel speed of individual vehicles passing through the scene units, and based on this result, calculate the average travel speed of the same type of vehicles passing through each scene unit.

[0063] Step 31) Calculate the travel time t(j) and average travel speed of individual vehicle j through scene unit i

[0064]

[0065] Among them, T O (j) is the recorded time when the vehicle arrives at the upstream node; T E (j) is the recorded time when the vehicle arrives at the downstream node; L is the driving distance of the vehicle in scene unit i.

[0066] Step 32) Calculate the average speed of vehicles of the same type passing through each scene unit

[0067]

[0068] Among them, n r,k is the number of vehicles passing through the ramp.

[0069] Step 4) Divide the dynamic time interval and calculate the ramp diversion coefficient.

[0070] Step 41) Considering the periodicity of traffic flow and the significant differences in traffic status in different time periods, a day is dynamically divided into ρ time intervals.

[0071] ρ t =Time(t,t+γ]t=0,1,…ρ

[0072] Among them, μ represents the time window, and Time(t,t+γ] represents the corresponding time interval. If the minimum time window is one hour, then γ=1, ρ=24, and the corresponding ρ 20 =Time(20,21], which means 20:00-21:00 in the evening.

[0073] Step 42) Calculate the ramp split coefficient

[0074] When estimating the traffic status in the detection blind spot, based on the matching of license plate information of each ETC gantry, the "source" and "destination" of the traffic flow are analyzed to restore the vehicle's driving route. Generally, the "source" of the downstream section traffic flow includes two places: the upstream ETC gantry observation station and the ramp entrance (from the ramp toll station or interchange). Similarly, the "destination" of the upstream traffic flow also includes two places: the downstream section or the exit ramp.

[0075] Based on the different scene units segmented in step 1), the exit ramp diversion coefficients R of different types of vehicles are calculated. off-ramp , entrance ramp split coefficient R on-ramp , combined entrance and exit ramp flow diversion coefficient R ramp The specific calculation is as follows:

[0076]

[0077]

[0078]

[0079] Among them, ρ t Indicates the time interval, N A,k (ρ t ) and N B,k (ρ t ) is the statistical time interval ρ t The collection of K-type vehicles passing through the upstream and downstream ETC gantries.

[0080] Step 5) Estimate the average travel time of the same type of vehicles passing through each scene unit in different dynamic time intervals.

[0081] Step 51) Calculate the k-type vehicle ρ on scene unit i t Average driving speed The specific calculation is as follows:

[0082] when hour:

[0083]

[0084] when hour:

[0085]

[0086] Among them, n i,k (ρ t ) is ρ t The number of k-type vehicles passing through section i is obtained by counting, is tThe βth scenario obtained by statistics is the set of k-type vehicles passing through road section i, and μ represents the fixed speed limit value on the special speed limit section.

[0087] Step 52) Based on the dynamic time interval divided in step 4) and the different scene units divided in step 1), estimate the average travel time of the k-type vehicle passing through the scene unit i The specific calculation is as follows:

[0088]

[0089] Among them, L i is the distance traveled by the vehicle in scene unit i.

[0090] Step 6) calculates the travel time for each type of vehicle to reach any target section downstream based on the average travel time obtained in step 5).

[0091] Taking any ETC gantry as the starting point, based on the average travel time in step 5) And the different scene units divided in step 1) calculate the travel time of the k-type vehicle to reach any target section D downstream The specific calculation is as follows:

[0092]

[0093] Among them, L D is the distance from the upstream node to the target location.

[0094] Step 7) Select a spatiotemporal prediction model to predict the fine-grained traffic flow of the ETC gantry in the future time period.

[0095] The prediction model can be selected based on the currently popular deep learning method. This embodiment selects a model based on STGCN improvement. The topological structure of the highway network is described by the graph structure G = (V, E, W), where all ETC gantries are the set of vertices V, E is the edge set composed of all roads, and W represents the adjacency matrix; the traffic flow and adjacency matrix aggregated every 5 minutes are used as input, and the model is used to predict and output the future specified time window (T demand ) of fine-grained traffic flow. The network framework of its prediction model, such as Figure 3 shown.

[0096] Step 8) Based on the ramp diversion coefficient, the travel time of each type of vehicle to reach any target section downstream, and the predicted fine-grained traffic flow of the ETC gantry in the future time period, the fine-grained traffic flow of any target section downstream and at any time is aggregated to obtain the traffic flow prediction value of the detection blind spot.

[0097] Step 81) Based on the different scene units segmented in step 1, the traffic estimation of the target location is focused into three scenarios, such as Figure 4 shown.

[0098] Step 82) Different estimated time requirements impose time constraints on the estimation method. Based on the vehicle arrival time calculated above make And considering different demand durations T demand , discussed under two conditions:

[0099] i)T demand ≤T traveling When the flow of k-type vehicles at the target location within the ΔT time period is counted,

[0100] ii) T demand >T traveling When the flow of k-type vehicles at the target location within the ΔT time period is predicted, The specific calculation is as follows:

[0101] Scenario 1: When β∈Scene{1,5}ori∈up_Seg ub (i), that is, the target position is on section L Location1 , then the target position in the future T demand The subsequent traffic is only related to the upstream ETC gantry:

[0102]

[0103]

[0104]

[0105] in, The traffic demand time is T demand The flow of k-type vehicles at the target location; is the distance from the current ΔT his The historical statistical values ​​of the upstream ETC gantry in the ΔT time period; For the future ΔT pre The predicted value of the upstream ETC gantry in the ΔT time period.

[0106] Scenario 2: When β∈Scene{3}andi∈down_Seg ub (i), that is, the target position is on section L Location2 , then the target position in the future T demand The traffic volume is related to the upstream ETC gantry and exit ramp:

[0107]

[0108]

[0109] Among them, R off-ramp is the k-type vehicle on road segment i in ρ t The diversion coefficient of the exit ramp traffic flow in the interval, The statistical time interval is ρ t The set of k-type vehicles exiting the ramp at time N A,k (ρ t ) and N B,k (ρ t ) is the statistical time interval ρ t The collection of K-type vehicles passing through the upstream and downstream ETC gantries.

[0110] Scenario 3: When β∈Scene{2}andi∈down_Seg ub (i) Then the target position in the future T demand The traffic volume is related to the upstream ETC gantry and entrance ramp:

[0111]

[0112]

[0113] Among them, R on-ramp is the k-type vehicle on road segment i in ρ t The traffic diversion coefficient of the entrance ramp of the interval, The statistical time interval is ρ t The set of k-type vehicles entering the ramp at time .

[0114] Scenario 4: When β∈Scene{4}andi∈down_Seg ub (i), that is, the target position is on section L Location3 , then the target position in the future T demand The traffic volume is related to the upstream ETC gantry and the entrance and exit ramps:

[0115]

[0116]

[0117] Among them, R ramp is the k-type vehicle on road segment i in ρ t The traffic diversion coefficient of the entrance and exit ramps of the interval.

[0118] Step 9) Design a verification scheme for detecting vehicle travel time and traffic flow in the blind spot, and evaluate the estimation performance of each variable in each verification scheme.

[0119] In order to explore the influence of spatial position on the traffic state estimation result, this embodiment designs four estimation schemes (estimation 1, 2, 3, 4), and selects ETC1-ETC4 to estimate the traffic state at the target position D1, such as Figure 5 shown.

[0120] Due to the driving time of vehicles in the detection blind spot and traffic flow It is impossible to measure directly, so a special evaluation scheme needs to be designed for the detection blind area. The present invention proposes to make the ETC gantry equivalent to the target position, use other ETC gantries to infer the traffic status of the equivalent position, and then verify the corresponding model estimation performance.

[0121] This embodiment designs four verification schemes (verification 1, 2, 3, 4), such as Figure 5 As shown, the ETC5 gantry is selected as the equivalent target position D2, and then the traffic state values ​​of ETC1-ETC4 are used to estimate the traffic state of the target position.

[0122] In summary, the present invention takes into account the spatial transfer effect and timeliness issues in the modeling process, realizes the fine-grained estimation task of highway traffic status, and thus improves the estimation effect.

[0123] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning, or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A fine-grained traffic state estimation method based on ETC gantry data, characterized in that: The following steps are involved: Step 1) Divide the road network into multiple scene units, divide the vehicle types into fine-grained types, and select vehicle travel time and traffic flow as traffic state variables; Step 2) Filter and process ETC gantry charging data to construct a fine-grained vehicle set that passes through each scene unit within a fixed time; Step 3) calculating the travel time and average travel speed of individual vehicles passing through the scene units, and based on the results, calculating the average travel speed of the same type of vehicles passing through each scene unit; Step 4) dividing the dynamic time interval and calculating the ramp diversion coefficient; Step 5) estimating the average travel time of the same type of vehicles passing through each scene unit in different dynamic time intervals; Step 6) calculating the travel time for each type of vehicle to reach any target section downstream based on the average travel time obtained in step 5); Step 7) Select a spatiotemporal prediction model to predict the fine-grained traffic flow of the ETC gantry in the future time period; Step 8) Based on the ramp diversion coefficient, the travel time of each type of vehicle to reach any target section downstream, and the predicted fine-grained traffic flow of the ETC gantry in the future time period, the fine-grained traffic flow of any target section downstream and at any time is aggregated to obtain the traffic flow prediction value of the detection blind spot.

2. According to claim 1, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: In the step 1), the segmentation rule for segmenting the road network into multiple scene units is: taking the complex road section locations that may affect the evolution of traffic flow as specific segmentation nodes, and connecting any two adjacent segmentation nodes as scene units, wherein the complex road section locations include ETC gantries, toll stations, ramps, tunnel entrances and exits, bridges, and ramps.

3. According to the method of claim 1, the fine-grained traffic state estimation method based on ETC gantry data is characterized in that: The step 2) is specifically as follows: Construct a set of k-type vehicles that pass through scene unit i in a fixed time in, n i,k is the statistic of k-type vehicles passing through scene unit i, veh represents the vehicle, and β represents the scene type.

4. According to claim 3, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: The step 3) is specifically as follows: Calculate the travel time t(j) and average travel speed of individual vehicle j through scene unit i Based on this result, calculate the average speed of vehicles of the same type passing through each scene unit. Among them, T O (j) is the recorded time when the vehicle arrives at the upstream node; T E (j) is the recorded time when the vehicle arrives at the downstream node; L is the distance traveled by the vehicle in scene unit i; n r,k is the statistics of k-type vehicles passing through the ramp; is the statistic of passing the ramp; n A,k is the statistics of k-type vehicles passing through the upstream ECT gantry, is the set of all k-type vehicles passing through the upstream ECT gantry obtained by statistics within a fixed time, up_Seg ub (i) down_Seg ub (i) Respectively represent the upstream and downstream sections without ETC.

5. According to claim 1, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: In step 4), taking into account the periodicity of traffic flow and the significant differences in traffic conditions in different time periods, one day is dynamically divided into p time intervals, and based on the matching of license plate information of each ETC gantry, the source and destination of the vehicle are analyzed, and the vehicle driving route is restored to accurately analyze the spatial propagation effect of the traffic flow.

6. According to the method of claim 1, the fine-grained traffic state estimation method based on ETC gantry data is characterized in that: In step 4), the ramp diversion coefficient is calculated by: based on the different scene units divided in step 1), the exit ramp diversion coefficients R of different types of vehicles are calculated. off-ramp , entrance ramp diversion coefficient R on-ramp and the combined entrance and exit ramp split coefficient R ramp ,in, Among them, ρ t Indicates the time interval, N A，k (ρ t ) and N B，k (ρ t ) is the statistical time interval ρ t The collection of K-type vehicles passing through the upstream and downstream ETC gantries.

7. According to claim 1, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: The step 5) is specifically as follows: based on the dynamic time interval divided in step 4), the average travel time of the k-type vehicle passing through the scene unit i is estimated in, represents the k-type car on scene unit i t The average driving speed of the time period is calculated according to different scene types; L i is the distance traveled by the vehicle in scene unit i.

8. According to claim 7, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: The step 6) is specifically as follows: taking any ETC gantry as the timing starting point, based on the average travel time of step 5) According to the scene types of different scene units, the travel time of the k-type vehicle to reach any downstream target section D is calculated.

9. According to claim 1, a fine-grained traffic state estimation method based on ETC gantry data is characterized in that: The step 8) is specifically as follows: Based on the travel time of a k-type vehicle to reach any downstream target section D make And determine the required time T demand , the traffic flow prediction value of the detection blind area is determined according to two conditions: i) When T demand ≤T traveling When the flow of k-type vehicles at the target location within the ΔT time period is determined based on historical statistical values ii) When T demand >T traveling When the fine-grained traffic flow of the predicted ETC gantry in the future time period is determined according to the predicted traffic flow of the k-type vehicle at the target position in the ΔT time period, 10. The fine-grained traffic state estimation method based on ETC gantry data according to claim 1 is characterized in that: The method further comprises: Step 9) Design a verification scheme for detecting vehicle travel time and traffic flow in the blind spot, and evaluate the estimation performance of each variable in each verification scheme. The verification scheme equates the ETC gantry to the target position, uses other ETC gantries to infer the traffic status of the equivalent target position, and evaluates the corresponding model estimation performance.

Images