Traffic flow prediction device

Abstract

Based on data characteristics inferred from past data, the system reduces processing load while suppressing a decline in prediction accuracy. [Solution] The traffic flow prediction device (10) acquires data on past dates and times and the current status of the moving object (20), determines the resolution of the data to be newly transmitted from the moving object (20) using the data analysis results or the traffic flow prediction results based on the data, and makes a prediction of future traffic flow using the data transmitted from the moving object (20) at that resolution.

Description

Technical Field

[0001] This disclosure relates to a traffic flow prediction device.

Background Art

[0002] In recent years, technologies for analyzing the flow of moving objects have been developed. For example, Patent Document 1 discloses a pedestrian flow analysis system that can analyze pedestrian flow data while suppressing an excessive increase in the processing load for analyzing the pedestrian flow data.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the technology described in Patent Document 1, the process for suppressing high load is only a cleansing process (only a process of thinning out position information indicating the same position) based on a uniform GPS-based standard, and does not reflect the trade-off regarding the estimation accuracy based on data characteristics inferred from past data.

[0005] In view of such circumstances, an object of the present disclosure is to be able to reduce the processing load while suppressing a decrease in prediction accuracy based on data characteristics inferred from past data.

Means for Solving the Problems

[0006] A traffic flow prediction device according to an embodiment of the present disclosure acquires past date and time and data of moving objects at the current time, determines the resolution of data newly transmitted from the moving objects using the analysis result of the data or the prediction result of the traffic flow based on the data, and predicts the future traffic flow using the data newly transmitted from the moving objects at the resolution. [Effects of the Invention]

[0007] According to this disclosure, it is possible to reduce the processing load while suppressing a decrease in prediction accuracy based on data characteristics inferred from past data. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows an example configuration of a traffic flow prediction system according to one embodiment. [Figure 2] This flowchart shows the processing procedure of a traffic flow prediction device according to one embodiment. [Figure 3] This is an illustrative diagram showing the difference in data resolution. [Modes for carrying out the invention]

[0009] The traffic flow prediction system 1 according to one embodiment shown in Figure 1 comprises one or more moving entities (vehicles, people, etc.) 20 and a traffic flow prediction device (data control device) 10. The traffic flow prediction device 10 may be a database server. The traffic flow prediction device 10 is installed in a center. The vehicles 20 and the traffic flow prediction device 10 are connected to each other via a network so that they can communicate with each other.

[0010] Each mobile unit 20 comprises one or more sensors and a communication unit (communication interface) for communicating with the traffic flow prediction device 10. Each mobile unit 20 transmits driving data detected by the sensors (hereinafter simply referred to as "data") to the traffic flow prediction device 10. The "data" includes, for example, vehicle identification number, GPS location information, time, vehicle speed, acceleration, driving operations (accelerator, brake, etc.), and attribute information of the driver and passengers.

[0011] The traffic flow prediction system 1 acquires past date and time data and current (real-time) data from moving objects 20 in the road traffic network as a moving data log. It then feeds back the spatiotemporal characteristics of the road traffic network obtained from the moving data log, the prediction results of traffic flow (traffic volume, passing speed, bias of attributes of moving objects 20, etc.) (results of the prediction task execution), etc. The traffic flow prediction system 1 considers the balancing of the amount of data used in the traffic flow prediction task and the prediction accuracy, and adaptively (dynamically) reduces the resolution of the data. As a result, the traffic flow prediction system 1 can reduce the amount of new data acquired from moving objects 20 while suppressing a decrease in prediction accuracy, thereby reducing the communication load and computation load. The traffic flow prediction system 1 can adaptively determine the resolution (sampling period of data to be acquired, group of moving objects to be acquired, etc.) for each situation such as the target road link and time of day. Furthermore, the resolution reduction process can be performed either as distributed processing on the edge side (mobile body 20 side) or as centralized processing on the center side (traffic flow prediction device 10 side).

[0012] Referring to Figure 2, an example of the processing procedure of the traffic flow prediction device 10 will be explained.

[0013] In S1, the traffic flow prediction device 10 determines the resolution for balancing processing based on past data. The traffic flow prediction device 10 dynamically determines the resolution specifications for newly transmitted data from the moving objects 20 using the analysis results of past data from the moving objects 20 or the traffic flow prediction results. For example, the traffic flow prediction device 10 may analyze past data from the moving objects 20 on the road network for each road link and time period, and determine the reduced resolution based on the spatiotemporal laws of traffic flow stability. Alternatively, the traffic flow prediction device 10 may determine the reduced resolution based on the traffic flow prediction results and empirically obtained from past prediction accuracy, within a range that maintains the desired prediction accuracy.

[0014] In S2, the traffic flow prediction device 10 determines whether or not to perform balancing processing on the edge side.

[0015] In S3, when the traffic flow prediction device 10 performs balancing processing on the edge side (Yes in S2), it instructs the moving body 20 of the resolution of the data newly transmitted from each moving body 20 (the resolution determined in S1 or S9) (edge side balancing processing).

[0016] In S4, the traffic flow prediction device 10 receives in real time data of the resolution determined in S3 from the moving body 20.

[0017] In S5, when the traffic flow prediction device 10 does not perform balancing processing on the edge side (No in S2), it receives in real time data of the default resolution from the moving body 20.

[0018] In S6, the traffic flow prediction device 10 changes the resolution of the data received in S5 according to the specification of the resolution determined in S1 and reduces the data volume (center side balancing processing). When the traffic flow prediction device 10 does not perform the edge side balancing processing in S3, it performs the center side balancing processing in S6.

[0019] In S7, the traffic flow prediction device 10 uses the data that has been balanced in S3 or S6 and transmitted from each moving body 20 at the specified resolution as input to predict the future traffic flow. In the first embodiment described later, an extrapolation of the total traffic volume is performed, and in the second embodiment, prediction of traffic statistical information is performed.

[0020] In S8, the traffic flow prediction device 10 determines whether to perform iterative processing. If the confidence level exceeds the target value or other cases where the iteration condition is not satisfied (No in S8), the processing ends.

[0021] In S9, when performing iterative processing (Yes in S8), the traffic flow prediction device 10 updates the specification of the resolution of the balancing processing according to the confidence level of the traffic flow result predicted in S7 and returns to the processing in S2.

[0022] (First Embodiment - Extrapolation of Total Traffic Volume -) The estimation of the expansion of the total traffic volume is a process of estimating the total traffic volume of the road traffic network based on data obtained only from some connected vehicle groups (for example, specific vehicle type groups, vehicle groups of specific manufacturers, etc.). It is known that the traffic volume of the vehicle group obtained from the location information, time information, etc. included in the data is one of the important explanatory variables (for example, see the following references). [References] Yong, J., Wakabayashi, Y., Okayasu, A., Miki, R., Sasai, T., Inoue, M., & Fukushima, S. (2022). Estimating Total Traffic Volume with Statistical Modeling Approach. In 2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC), pp. 304-309, IEEE.

[0023] In S1, the traffic flow prediction device 10 determines a combination of situation and granularity from past data based on a preselected criterion. That is, the traffic flow prediction device 10 reduces the resolution of the data by making the granularity as coarse as possible for each situation, making the data easier to handle quantitatively.

[0024] Figure 3 is an image diagram showing the process of S1. Near the raw data resolution is used for links with unstable traffic flow, and the resolution is coarsened (the reduction rate is increased) from the raw data for links with stable traffic flow. The resolution specification is composed of a combination of situation and granularity and is determined to conform to the criterion. Hereinafter, the details of the setting of the criterion, situation, and granularity will be described.

[0025] The "criterion" regarding the determination of the situation and granularity related to the resolution is determined according to the characteristics of the past data. Examples of the determination criteria used for data volume reduction are given below. (1) Utilize the spatio-temporal regularity regarding the stability of past data For example, the criteria for selection include traffic volume (locations and times with low traffic dispersion, locations with strong periodicity between time of day and day of the week, etc.), the tendency for sudden events to occur (locations and times where sudden events such as accidents, congestion, and traffic restrictions are unlikely to occur, etc.), and the bias in the attributes of the mobile entities 20 (locations with a high penetration rate of the vehicle group to be acquired). (2) Use the past performance of the accuracy of estimates of overall traffic volume. For example, based on past experience, the location and time where the prediction accuracy of augmented estimation is high even with a small amount of data can be determined as the criteria.

[0026] The "situation" for setting the granularity can be broadly classified into two categories: time and space. (i) Temporal circumstances For example, the granularity can be determined for each unit of time of day, day of the week, or season. (ii) Spatial situation For example, the level of detail is determined for each region and road unit.

[0027] The "granularity" specifications for newly acquired data can be broadly categorized and set into two types depending on the situation: time granularity and sample granularity. (a) Time granularity (sampling period) For example, the sampling period can be changed depending on the situation. (b) Specimen particle size For example, the proportion of vehicles selected for sampling from among the vehicles from which data can be acquired can be changed depending on the situation. For example, the number of data items (vehicle speed, fuel consumption, etc.) selected for sampling can be changed depending on the situation.

[0028] The traffic flow prediction device 10, for example, combines (2)-(i)-(a) above to simulate a lower resolution sampling period for all vehicle data on expressways in Tokyo over the past 24 hours, and determines a threshold sampling period for each time unit in predetermined time intervals (e.g., 1 hour) where the average error of the magnified estimation is less than a predetermined value (e.g., 10%), and sets this threshold as the resolution of the newly transmitted data.

[0029] Furthermore, the traffic flow prediction device 10 combines, for example, (1)-(ii)-(b) above to determine the rate at which to instruct data acquisition to be performed on the group of vehicles to be targeted for data acquisition, according to the penetration rate (coverage rate) of vehicles in each region. If it is sufficient to know 25% of all passing vehicles, then on roads in areas where the penetration rate of the target vehicles is 50%, data transmission is instructed to 50% of those vehicles, and on roads in areas where the penetration rate of the target vehicles is 25%, data transmission is instructed to 100% of those vehicles.

[0030] In S3, the traffic flow prediction device 10 transmits the resolution specifications determined in S1 to the target vehicles on the road traffic network that are to acquire data. In subsequent data transmissions, the target vehicles transmit data at the resolution specified in the specifications, resulting in lower data resolution and reduced communication load.

[0031] In S4 and S5, the traffic flow prediction device 10 acquires data in real time from the target vehicle group for data acquisition. In S4, the traffic flow prediction device 10 acquires data at different resolutions depending on the situation, such as every 300 milliseconds for road A on weekdays, every 400 milliseconds for road A on holidays, every 200 milliseconds for road B on weekdays, and every 500 milliseconds for road B on holidays. In S5, the traffic flow prediction device 10 acquires data at a 500-millisecond interval for all roads at all times, using the default setting.

[0032] In S6, the traffic flow prediction device 10 performs a process to reduce the resolution of the data received in S5 according to the resolution specifications determined in S1.

[0033] In S7, the traffic flow prediction device 10 performs an expanded estimation of the total traffic volume using the data after balancing. For example, the traffic flow prediction device 10 learns the parameters of a regression model (linear regression model, mixed-effects model, etc.) from past data, then inputs newly acquired real-time data and outputs the prediction result. For example, the dependent variable is the total traffic volume per hour for each road, and the independent variables are the traffic volume of vehicles targeted for data acquisition per hour for each road, the average vehicle speed, the specifications of the target road, the number of lanes, and the time of day / day of the week category. The features may be those obtained directly from the data (traffic volume per road for a group of vehicles, time of day, day of the week, presence or absence of holidays, etc.) or those obtained in combination with information from an external database (road characteristics of the road where the vehicle is located (road specifications, number of lanes, etc.), weather conditions, etc.).

[0034] In S8, the traffic flow prediction device 10 checks the degree of fit of the data after balancing to a traffic engineering model (such as a QV curve) calculated in advance from past data, and calculates the confidence level of the data for each resolution. The traffic flow prediction device 10 sets the repeated determination to Yes only if the confidence level of the obtained data is below a predetermined threshold and it is possible to increase the resolution of the acquired data (the current resolution specification setting is less than the system's upper limit).

[0035] In S9, the traffic flow prediction device 10 updates the specification of the granularity of the situation at a resolution where further data acquisition is deemed necessary, in the same manner as in S1, according to the confidence level of the repeated determination.

[0036] (Second Embodiment - Prediction of Traffic Statistics Information) The system performs balancing processing on data from moving objects 20 used for prediction tasks related to statistical quantities such as average speed and average travel time (travel time) for road-specific traffic statistics information (statistics related to the smoothness of traffic flow) in the road traffic network. Differences from the first embodiment are indicated below by enclosing them in <>.

[0037] In S1, the resolution specification for balancing processing is determined by a combination of situation and granularity from past data based on pre-selected <decision> criteria. The traffic flow prediction device 10, for example, simulates reducing the resolution of the sampling period for all vehicle data for the past 24 hours on expressways in Tokyo, and simulates a threshold sampling period at each time in predetermined time units (e.g., 1 hour) where the average error of <average speed estimation or average passage time> is less than a predetermined value (e.g., 10%), and determines this threshold as the resolution of the newly transmitted data.

[0038] In S7, the traffic flow prediction device 10 performs <estimate of traffic statistics information> using the data after balancing. For example, the traffic flow prediction device 10 learns the parameters of a regression model from past data, then inputs newly acquired real-time data and outputs the prediction result. For example, the dependent variable is the <average vehicle speed> per hour for each road, and the independent variables are the <average vehicle speed, traffic volume> of the target vehicles per hour for which data is acquired, the specifications of the target road, the number of lanes, and the time of day / day of the week category. Features may be obtained directly from the data (e.g., <average vehicle speed> per road for the vehicle group, time of day, day of the week, presence or absence of holidays), or they may be obtained in combination with information from an external database.

[0039] In the above embodiment, if the moving body 20 is specifically a "vehicle," in S1, the traffic flow prediction device 10 may determine the resolution by considering the characteristics of the vehicle data, including driving characteristics and vehicle characteristics, as well as the characteristics of road traffic, including road characteristics and traffic characteristics. "Driving characteristics" are characteristics estimated from past data using a driver model, representing biases in driving operations (quality of driving skills, frequency of detours and breaks, etc.). "Vehicle characteristics" are the driving performance of the vehicle, estimated from past data such as vehicle model, total mileage, and fuel consumption. "Road characteristics" are information on road links related to the smoothness of traffic flow (number of lanes, width, curvature, etc.). "Traffic characteristics" are traffic regulation information related to the Road Traffic Act (speed limits, regulatory signs, etc.), traffic information (congestion and accident information provided by operators, etc.), etc. Furthermore, the traffic flow prediction device 10 may use a degree of fit utilizing the above characteristics to determine whether or not to perform the repeated processing in S8.

[0040] Since the data from the mobile object 20 is spatiotemporally continuous, if data resolution is prioritized and sampling is continued at minute time intervals or large sample sizes, the amount of data becomes enormous, resulting in an unrealistic load on communication and the computing load for task processing. In this regard, the traffic flow prediction device 10 determines the resolution of the data newly transmitted from the mobile object 20 using the analysis results of the data acquired from the mobile object 20 or the traffic flow prediction results, and uses the newly transmitted data at that resolution to predict future traffic flow. As a result, the traffic flow prediction device 10 can reduce the processing load while suppressing a decrease in prediction accuracy.

[0041] To enable the aforementioned traffic flow prediction device 10 to function, a computer capable of executing program instructions can also be used. The program may be recorded on a non-transitory recording medium that the computer can read.

[0042] Although the embodiments described above are representative examples, it will be apparent to those skilled in the art that modifications and substitutions can be made within the spirit and scope of the present invention. For example, it is possible to combine multiple steps described in the flowchart of the embodiments into one, or to divide a single step. [Explanation of symbols]

[0043] 1: Traffic flow prediction system, 10: Traffic flow prediction device, 20: Mobile body

Claims

1. By obtaining data on past dates and times and the current state of movement, Using the analysis results of the aforementioned data, or the traffic flow prediction results based on the aforementioned data, the resolution of the data newly transmitted from the mobile body is determined. A traffic flow prediction device that predicts future traffic flow using data newly transmitted from the mobile body at the aforementioned resolution.

2. The aforementioned traffic flow forecast is an estimate of the overall traffic volume, The traffic flow prediction device according to claim 1, wherein the acquired data is simulated to have a lower resolution, a resolution threshold at which the average error of the traffic flow prediction result is less than a predetermined value is determined in predetermined time units, and this threshold is determined to be the resolution of the newly transmitted data.

3. The aforementioned traffic flow prediction is based on traffic statistics information. The traffic flow prediction device according to claim 1, wherein the acquired data is simulated to have its resolution reduced, a resolution threshold is determined in predetermined time units at which the average error in the average speed estimation or average passage time is less than a predetermined value, and this threshold is determined to be the resolution of the newly transmitted data.

Images