Highway service area vehicle residence time estimation method based on kinematics principle

Abstract

The application discloses a highway service area vehicle residence duration estimation method based on kinematic principle, which converts the vehicle residence duration estimation problem into the vehicle actual driving duration estimation problem based on the kinematic principle. The experience of the service area residence vehicle is divided into five processes, i.e. the upstream stable driving stage, the service area entrance ramp slow deceleration stage, the service area residence and rest stage, the service area exit ramp acceleration driving out stage and the downstream stable driving stage, and the driving duration of each stage except the service area residence and rest stage is calculated respectively. Finally, the residence duration of the service area is obtained. Compared with the traditional average speed method and the commonly used machine learning model, the calculation accuracy of the application is higher, and the robustness of the application is stronger in the root mean square error (RMSE).

Description

A method for estimating vehicle dwell time at highway service areas based on kinematic principles Technical Field

[0001] This invention relates to the field of highway service technology, and in particular to a method for estimating vehicle dwell time in highway service areas based on kinematic principles. Background Technology

[0002] Currently, vehicle information collection in service areas mainly relies on checkpoint cameras. Some large service areas are attempting to enhance the dimensionality and quality of supplementary data through equipment such as radar and radar-video fusion systems. However, existing technical solutions have their own limitations, resulting in the accuracy, completeness, and stability of the collected vehicle information failing to meet the requirements for effective analysis of service area operations. Summary of the Invention

[0003] The purpose of this invention is to provide a method for estimating vehicle dwell time in highway service areas based on kinematic principles.

[0004] The technical solution adopted in this invention is:

[0005] A method for estimating vehicle dwell time at highway service areas based on kinematic principles includes the following steps:

[0006] Step 1: Obtain road data for the service area and data from two ETC gantries before and after the service area;

[0007] Step 2: Calculate the actual travel time Δt of the service area based on the entrance and exit gantry data. QD2 ;

[0008] Step 3: Calculate the average speed v1 of the previous section of the service area and use it as the driving speed in the smooth section from the service area entrance gantry to the service area entrance ramp divergence point.

[0009] Step 4: Calculate the travel time within the entrance smooth zone based on the distance Δs1 of the entrance smooth zone.

[0010] Step 5: Calculate the travel time at the service area entrance ramp. Δs2 is the distance to the service area entrance ramp. This refers to the driving speed at the exit of the entrance ramp;

[0011] Step 6: Calculate the average speed v3 of the next section after the service area, and use it as the driving speed in the smooth section from the merging point of the service area exit ramp to the exit gantry of the service area exit.

[0012] Step 7: Calculate the travel time to the service area exit ramp. v 30 Let a be the initial velocity at the entrance of the exit ramp. + This refers to the vehicle's acceleration during the acceleration phase at the exit ramp.

[0013] Step 8: Calculate the travel time in the exit stability zone. Δs5=Δs s -Δs4, Δs s The distance from the service area to the third gantry;

[0014] Step 9: Finally, calculate the vehicle's dwell time Δt at the service area. s

[0015]

[0016] Furthermore, the road data for the service area includes the distance from the service area entrance gantry to the smooth section of the service area entrance ramp, the distance of the entrance ramp, the distance of the exit ramp, and the distance from the service area exit ramp to the smooth section of the service area exit gantry.

[0017] Furthermore, the actual travel time Δt in the service area QD2 Including the duration of stay at the service area Δt s Travel time Δt to service area r Travel time to service areas

[0018] Furthermore, without loss of generality, let the initial velocity v at the entrance of the exit ramp be... 30 and the driving speed at the entrance ramp exit If the value is 0, then the dwell time in the service area is...

[0019] Furthermore, a + The value range is 0.8–1.2 m·s -2 .

[0020] This invention employs the above technical solution, based on a kinematic principle-based method for estimating vehicle dwell time, transforming the problem of estimating vehicle dwell time into the problem of estimating actual vehicle travel time. The experience of a vehicle parked at a service area is divided into five stages: an upstream steady travel stage, a slow deceleration stage at the service area entrance ramp, a service area rest and adjustment stage, a service area exit ramp acceleration and exit stage, and a downstream steady travel stage. The travel time for each stage, excluding the service area rest and adjustment stage, is calculated. Finally, the dwell time at the service area is obtained. Compared to traditional average speed methods and commonly used machine learning models, this invention offers higher computational accuracy and stronger robustness in terms of root mean square error (RMSE). Attached Figure Description

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0022] Figure 1 is a schematic diagram of the five stages of vehicle parking in the service area;

[0023] Figure 2 shows the distribution of vehicle dwell time in service area A.

[0024] Figure 3 shows the distribution of vehicle dwell time in service area B.

[0025] Figure 4 is a schematic diagram comparing the cumulative probability of estimation errors in service areas A and B. Detailed Implementation Methods

[0026] 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.

[0027] As shown in any one of Figures 1 to 4, this invention discloses a method for estimating vehicle dwell time in highway service areas based on kinematic principles, which includes the following steps:

[0028] Step 1: Obtain road data for the service area and data from two ETC gantries before and after the service area. The road data for the service area includes the distance from the service area entrance gantry to the smooth section of the entrance ramp, the distance of the entrance ramp, the distance of the exit ramp, and the distance from the service area exit ramp to the smooth section of the exit gantry.

[0029] Step 2: Calculate the actual travel time Δt of the service area based on the entrance and exit gantry data. QD2 The actual travel time Δt at the service area QD2 Including the duration of stay at the service area Δt s Travel time Δt to service area r Travel time to service areas

[0030] Step 3: Calculate the average speed v1 of the previous section of the service area and use it as the driving speed in the smooth section from the service area entrance gantry to the service area entrance ramp divergence point.

[0031] Step 4: Calculate the travel time within the entrance smooth zone based on the distance Δs1 of the entrance smooth zone.

[0032] Step 5: Calculate the travel time at the service area entrance ramp. Δs2 is the distance to the service area entrance ramp. This refers to the driving speed at the exit of the entrance ramp;

[0033] Step 6: Calculate the average speed v3 of the next section after the service area, and use it as the driving speed in the smooth section from the merging point of the service area exit ramp to the exit gantry of the service area exit.

[0034] Step 7: Calculate the travel time to the service area exit ramp. v 30 Let a be the initial velocity at the entrance of the exit ramp. + This refers to the vehicle's acceleration during the acceleration phase at the exit ramp.

[0035] Step 8: Calculate the travel time in the exit stability zone. Δs5=Δs s -Δs4, Δs s The distance from the service area to the third gantry;

[0036] Step 9: Finally, calculate the vehicle's dwell time Δt at the service area. s

[0037]

[0038] Furthermore, without loss of generality, let the initial velocity v at the entrance of the exit ramp be... 30 and the driving speed at the entrance ramp exit If the value is 0, then the dwell time in the service area is... a + The value range is 0.8–1.2 m·s -2 .

[0039] This invention employs the above technical solution, based on a kinematic principle-based method for estimating vehicle dwell time, transforming the problem of estimating vehicle dwell time into the problem of estimating actual vehicle travel time. The experience of a vehicle parked at a service area is divided into five processes: an upstream steady travel phase, a slow deceleration phase at the service area entrance ramp, a rest and adjustment phase at the service area, an acceleration phase at the service area exit ramp, and a downstream steady travel phase. The travel time for each process, excluding the rest and adjustment phase, is calculated. Finally, the dwell time at the service area is obtained.

[0040] The working principle of this invention will be explained in detail below:

[0041] The dataset of this invention includes an ETC transaction dataset and a service area traffic flow dataset. The ETC transaction dataset was collected from over 1,000 ETC gantries deployed across the entire expressway network in Fujian Province, from September 3rd to 10th, 2020. A total of 42,964,489 expressway ETC transaction data entries were obtained, including fields such as vehicle identifier, transaction time, gantry number, and trip number after anonymization, as shown in Table 1. Specifically, each transaction data entry contains all field information. Based on expressway vehicle type classification and toll standards, vehicles can be divided into 4 categories of passenger vehicles, 6 categories of freight vehicles, and 6 categories of special-purpose vehicles, totaling approximately 1.72 million vehicles.

[0042] Table 1 Description of some fields in ETC transaction data

[0043] Serial Number | Field Name | Field Attribute | Example 1. VehID | Vehicle Identifier | A000001 2. VehClass | Vehicle Type | 1 3. EnWeight | Total Axle Weight at Entrance | 1500 4. EnStation | Toll Station Entrance Number | 1002 5. EnTime | Toll Station Entrance Time | 2020 / 9 / 5 00:00:00 6. GantryID | Gantry Number | G00************0020 7. TradeTime | Transaction Time | 2020 / 9 / 5 01:00:00 8. Holiday | Statutory Holidays | 0 surface

[0044] The service area traffic flow experimental dataset was collected from the entrance and exit cameras of the Minhou Yangli Service Area (sections A and B) located at K1845 on the Beijing-Taipei Expressway in Fujian Province. The data collection period coincided with the ETC data collection period, yielding over 30,000 service area traffic flow data entries. These entries include fields such as service area number, de-identified vehicle identifier, capture time, and vehicle type, as shown in Table 2. The vehicles include four categories: passenger cars, buses, light trucks, and heavy trucks, totaling approximately 18,000 vehicles. It is important to note that this dataset was only used for experimental verification to evaluate the service area vehicle identification performance and the accuracy of vehicle dwell time estimation.

[0045] Table 2 Description of Service Area Traffic Flow Data Fields

[0046] Serial Number | Field Name | Field Attribute Example 1. SAID | Service Area Number | Yangli Service Area | Area A 2. EnEx | Entrance / Exit | 0 / 1 3. VehID | Vehicle Identifier | A000001 4. CapTime | Capture Time | 2020 / 9 / 5 00:00:00 surface

[0047] This invention only requires data from the service area and the two ETC gantries before and after it. Specifically, the deployment locations of the Yangli service area and its adjacent ETC gantries are shown in Figure 2. Gantry 1 and gantry 2 constitute segment 1, gantry 2 and gantry 3 constitute segment 2 (the service area segment), and gantry 3 and gantry 4 constitute segment 3. Therefore, it is necessary to process the discrete ETC transaction data into vehicle trajectory data and match and fuse the service area data.

[0048] After identifying vehicles parked at highway service areas, it is necessary to further estimate their dwell time (hereinafter referred to as: vehicle dwell time). Since the service area is located between the 2nd and 3rd gantries (segment 2), the total travel time in this segment consists of the actual travel time of the vehicles within the segment and their dwell time. Therefore, the vehicle dwell time Δt can be obtained. s :

[0049] Δt s =Δt QD2 -Δt r (19)

[0050] In the formula, Δt QD2 =tr 3.T -tr 2.T , Δt r This refers to the actual driving time of the vehicle to be determined.

[0051] Therefore, the problem of estimating vehicle dwell time is transformed into the problem of estimating actual vehicle travel time. Since highway traffic conditions are generally good, under non-congested and unforeseen circumstances, highways can approximate a free-flowing state, and vehicles typically travel relatively smoothly. Therefore, the average speed of the preceding and following sections can be used to estimate the actual vehicle travel time.

[0052]

[0053] Substituting formula (20) into formula (19), we get:

[0054]

[0055] Although the model is simple and easy to understand, it does not take into account the kinematic principles of vehicles entering / exiting highway service area ramps. Generally, vehicles parked in service areas go through five stages: upstream steady driving stage, slow deceleration stage at the service area entrance ramp, rest and adjustment stage at the service area, acceleration stage at the service area exit ramp, and downstream steady driving stage, as shown in Figure 1.

[0056] To address this, a vehicle dwell time estimation model based on kinematic principles was constructed, where the actual vehicle driving time Δt is the driving time. r for:

[0057] Δt r =Δt1+Δt2+Δt4+Δt5(22)

[0058] In the formula, Δt1 is the time taken during the upstream smooth driving phase; Δt2 is the time taken during the slow deceleration phase at the service area entrance ramp; Δt4 is the time taken during the acceleration phase at the service area exit ramp; and Δt5 is the time taken during the downstream smooth driving phase.

[0059] Phase 1, Upstream Stable Driving Phase: Based on the principle of inertia, the driving state in this phase can be considered a continuation of the driving state in the previous section, and can be approximated as uniform motion. Therefore, the average driving speed of the section before the service area can be used as the driving speed for this phase, and the time taken for the upstream stable driving phase can be obtained:

[0060]

[0061] In the formula, Δs1 is the distance from the second gantry to the entrance ramp divergence point; v1 is the average driving speed of the section before the service area.

[0062] Phase Two: Service Area Entrance Ramp Entry Stage: To some extent, highway service area ramps are similar to highway tollbooth entrance and exit ramps. However, tollbooth entrance and exit ramps are typically designed with high curvature, while service area ramps generally have low curvature, or even resemble straight lines, making the vehicle entry / exit process smoother. Therefore, the deceleration process at the service area entrance ramp can be approximated as uniformly decelerated linear motion. From Phase One, we know that the initial velocity of the uniformly decelerated linear motion is v1, the displacement is Δs2, and the velocity at time Δt2 is... and acceleration a - We can obtain:

[0063]

[0064]

[0065] Combining formulas (24) and (25), we get:

[0066]

[0067] In the formula, Δs2 is the distance from the service area entrance ramp divergence point to the service area.

[0068] Phase Four: Service Area Exit Ramp Exit Phase: Similarly, the acceleration process at the service area exit ramp can be approximated as uniformly accelerated linear motion until the speed reaches a steady state. From Phase Four, we know that the vehicle reaches a steady state after time Δt4, with a speed of v3. Also, assume the initial velocity of the uniformly accelerated linear motion is v... 30 and acceleration a + From formula (24), we can obtain:

[0069]

[0070] Under normal circumstances, a + =0.8~1.2m·s -2 .

[0071] Phase 5, Downstream Stable Driving Phase: Similar to the upstream stable driving phase, the driving state of the next segment can be considered a continuation of the driving state of this phase. Therefore, the average driving speed of the segment after the service area can be used as the driving speed of this phase, and the time taken for the downstream stable driving phase can be obtained:

[0072]

[0073] In the formula, v3 is the average driving speed of the section after the service area; Δs5=Δs s -Δs4, Δs s This refers to the distance from the service area to the third gantry.

[0074] Substituting formulas (23) and (26) to (28) into formula (22), we get:

[0075]

[0076] After processing, the mathematical model for estimating vehicle dwell time based on kinematic principles is as follows:

[0077]

[0078] Generally, the velocity at time Δt2 in uniformly decelerated linear motion can be considered as... The initial velocity v of uniformly accelerated linear motion 30 If all are 0, it can be simplified to:

[0079]

[0080] Statistical analysis of vehicle dwell time: Dwell time was segmented and statistically analyzed in 5-minute increments. The distribution of vehicle dwell time is shown in Figure 2 or 3. Figures 2 and 3 show that the dwell time in both areas A and B of Yangli Service Area exhibits a long-tail distribution. Furthermore, the largest number of vehicles have a dwell time of 5-10 minutes, and over 90% of vehicles do not stay in the service area for more than one hour. Specifically, the average dwell time in both areas A and B of Yangli Service Area is approximately half an hour, with a standard deviation of about 70 minutes. The longest dwell time exceeds 12 hours, while the shortest dwell time is less than 30 seconds.

[0081] Vehicle dwell time estimation accuracy: To evaluate the accuracy of vehicle dwell time estimation, the root mean square error (RMSE), mean absolute error (MAE), and R-coefficient are used to quantify the estimation error. The error is calculated as follows:

[0082]

[0083]

[0084]

[0085] In the formula, y represents the estimated vehicle dwell time obtained using the model. i This indicates the actual duration of vehicle dwell time. The average vehicle dwell time is represented by n, which represents the amount of data.

[0086] The proposed vehicle dwell time estimation method based on kinematic principles is compared and analyzed with the traditional average speed method and commonly used machine learning models. The experimental results are shown in Table 3. The experimental results show that the proposed method performs best, followed by the average speed method, while the machine learning model performs worst. Specifically, taking region B as an example, the average absolute error (MAE) of the proposed method is only 14s, which is more than double that of the average speed method and at least four times that of the machine learning method, indicating that the proposed method has higher accuracy. In terms of root mean square error (RMSE), the proposed method improves by at least one order of magnitude compared to the machine learning method, indicating stronger robustness. Furthermore, among the machine learning methods, XGBoost, RF, and GBDT are in the first tier, although they perform poorly in MAE and RMSE... 2While these models performed well on the MAE, their RMSE performance was only average, indicating that their fit to the data needs improvement. Models like Lasso, SVR, and KNN achieved extremely poor estimation results on all evaluation metrics, suggesting they are unsuitable for estimating vehicle dwell time. Notably, compared to XGBoost, RF, and GBDT, the average speed method, while achieving similar results on MAE, improved RMSE by nearly two times.

[0087] Table 3 Comparison of Estimated Service Area Dwell Time (Unit: seconds)

[0088]

[0089] Estimation Error Analysis: To further investigate the estimation error, a statistical analysis was performed on the mean absolute error of vehicle dwell time, and the cumulative probability distribution was plotted, as shown in Figure 4. The figure shows that the cumulative probability distribution curves in areas A and B both exhibit a rapid increase with the increase in the vehicle dwell time estimation error, eventually stabilizing after 2 minutes. Specifically, P{MAE≤120s}>95%, indicating that the probability of controlling the vehicle dwell time estimation error within 2 minutes exceeds 95%. In particular, the probabilities of controlling the estimation error in area B within 1 minute and 2 minutes are respectively P0 and P2. B {MAE≤60s}>97% and P B The MAE ≤ 120s > 99.8%, further verifying that the model has strong robustness.

[0090] Compared with the traditional average velocity method and commonly used machine learning models, the present invention has higher computational accuracy; and it is more robust to the root mean square error (RMSE).

[0091] Obviously, the described embodiments are only a portion, not all, of the embodiments of this application. Without conflict, the embodiments and features described and illustrated herein can be combined with each other. The components of the embodiments of this application generally described and illustrated in the accompanying drawings can 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 the 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 method for estimating vehicle dwell time at highway service areas based on kinematic principles, characterized by: It includes the following steps: Step 1: Obtain road data and ETC gantry data for the service area, including data from the two ETC gantries before and after the service area; Step 2: Calculate the actual travel time to the service area based on the entrance and exit gantry data. Step 3: Calculate the average speed of the previous segment of the service area. And as the driving speed of the smooth entry section from the service area entrance gantry to the service area entrance ramp divergence point; Step 4, based on the distance length of the smooth entry section The travel time in the inlet steady zone was calculated. Step 5: Calculate the travel time to the service area entrance ramp. ， This refers to the distance between the service area entrance ramps. Step 6: Calculate the average speed at the exit of the entrance ramp; Step 7: Calculate the average speed of the section following the service area. Step 7: Calculate the travel time on the service area exit ramp, and use this as the speed during the smooth exit section from the service area exit ramp merging point to the service area exit gantry; ， The initial velocity at the entrance of the exit ramp. Step 8: Calculate the vehicle's acceleration during the acceleration phase at the exit ramp; Step 9: Calculate the travel time in the exit steady zone. ， ， ， The distance from the service area to the third gantry; Step 9, finally calculate the vehicle's dwell time at the service area. ； 。 2. The method for estimating vehicle dwell time in highway service areas based on kinematic principles according to claim 1, characterized in that: The road data for the service area includes the distance from the service area entrance gantry to the smooth section of the service area entrance ramp, the distance of the entrance ramp, the distance of the exit ramp, and the distance from the service area exit ramp to the smooth section of the service area exit gantry.

3. The method for estimating vehicle dwell time in highway service areas based on kinematic principles according to claim 1, characterized in that: Actual travel time at the service area Including the length of stay in service areas Travel time to service area Travel time to service areas 。 4. The method for estimating vehicle dwell time in highway service areas based on kinematic principles according to claim 1, characterized in that: Set the initial velocity at the exit ramp entrance and the driving speed at the entrance ramp exit If the value is 0, then the dwell time in the service area is... 。 5. The method for estimating vehicle dwell time in highway service areas based on kinematic principles according to claim 1, characterized in that: The range of values ​​is 。

Images