Data-Driven Analysis for Road Traffic Conditions Using Digital Tachograph Data ^†

Abstract

Traffic condition analysis requires various conditions to be met using the conventional method. This has the limitation that it does not reflect the congestion caused by the actual vehicle flow. In this study, we suggest data-based traffic condition analysis. This is a method of determining the dynamic traffic conditions using Digital TachoGraph data, which can reflect the flow rate and the traveling speed of vehicles by the actual time zone. The suggested system could be implemented in both public and private sectors to create new possibilities and insights. For the public sector, police patrol vehicles can implement the system to create a ‘dynamic service area’ which would enhance efficiency for security patrols. For private sectors, the system could be applied to various call-dispatch systems to minimize waiting time for customers and driving distance for drivers. Also, it could be applied to upcoming autonomous vehicle sharing systems to ensure maximum coverage for autonomous cars.

1. Introduction

Conventional methods of traffic condition analysis most frequently rely on traffic flow analysis and traffic system analysis [1] in order to derive the traffic conditions, determining whether the traffic flow will increase or decrease at a certain location at a certain time. The conventional method is effective in terms of city planning and deriving various methods to control the flow of traffic within a city. However, since the conventional method uses systematic analysis of predicted values, it lacks the ability to reflect accidental events or real-world traffic conditions which could vary due to weather conditions or events taking place within a city.

In this study, we suggest a data-driven method to analyze traffic conditions using DTG (Digital TachoGraph) data within Seoul, South Korea. DTG data containing GPS trajectories of 30,000 vehicles within South Korea were used to create a serviceability map for the Seoul area. The map dynamically changed over time showing differences in the serviceable area within certain hours of the day. The serviceability map was drawn using the DTG data with GPS locations by calculating the movement of each vehicle within the given threshold time.

Using the suggested data-driven method could provide various aids for analyzing traffic conditions since it could reflect the actual movement of vehicles, which would react to real-world events such as accidents or the gathering of mass crowds.

2. Data

2.1. DTG Data

For the study, DTG data from January 2020 to December 2020 acquired from the Ministry of Land, Infrastructure and Transport (MOLIT) [2] were used. Information about the data is stated in

Table 1. Data dictionary (DTG data).

Column Name	Description	Example
Date	YYYY/MM/DD	20200104
Time	HH/MM/SS	123002
CARID	Encrypted car ID	A64229AA508BDDCAAE6B9923481C3442
CC	ID for vehicle company	1252361
DTD	Daily travel distance (km)	47
ATD	Accumulated travel distance of vehicle (km)	42,526
TS	km/h (000~255)	69
ACCX	Acceleration for X axis m/s2	−1.0
ACCY	Acceleration for Y axis m/s2	5.2
RPM	Rev per minute	1480
BS	0 (off) or 1 (on)	0
X	X GPS coordinate of vehicle (WGS84)	127.188979
Y	Y GPS coordinate of vehicle (WGS84)	36.919575
AZI	GPS azimuth in degrees (0~360)	179

. The size of the data is roughly 800 GB per month, since the DTG sensors create 1 row of data every 10 s. The total size of the data for 12 months exceeded 2.5 TB, thus preprocessing steps to remove unnecessary data were conducted beforehand.

DTG collects data from ‘company owned vehicles’, such as taxis, buses, freight trucks, etc., which requires continuous management. Originally the data were used for management purposes; however, since the data contain GPS information, for this study, the vehicle data were used as ‘agents’ depicting the traffic conditions of the Seoul area.

2.2. DTG Data Pre-Processing

The data contained the DTG information of about 30,000 vehicles which travel around South Korea. However, since special vehicles such as freight trucks which usually transport heavy loads throughout the whole of South Korea, using data collected from special vehicles could lead to biased results due to the specific usage of the vehicle. Hence, by using the company ID column, vehicles with specific usages were removed from the study data. Afterwards, the data were processed to only contain information within Seoul, the study area. The preprocessed data only contained vehicles within Seoul which includes in-town buses and private taxis.

3. Method

Service Area Analysis and Data-Driven Method

In order to extract traffic conditions using DTG data containing GPS information, the service area analysis method [3] was used. The conventional method for service area analysis uses the road network as baseline data and implements the Dijkstra algorithm to calculate the reachable distance within a given threshold. However, this method has limitations as it cannot reflect real-world conditions such as rush-hour congestion, resulting in a smaller service area within the same threshold time. Even though the conventional method can use average or maximum speed as inputs for each road segment, it only allows for a fixed value for speed and cannot consider real-world traffic conditions.

On the contrary, the suggested method of data-driven service area analysis uses accumulated GPS location data, which contains actual movement of vehicles in a specified time. By using actual vehicle GPS data, the traffic conditions can be reflected into the service area analysis, allowing for a more accurate representation of the service area size. For example, in late-night hours when traffic flow is low, vehicles tend to travel quicker than usual, while during rush hour with high traffic flow, vehicles will move much slower, providing a smaller service area.

4. Result

The result of the service area analysis within the Seoul area is depicted in Figure 1, where the green areas are areas which could be accessed within a 5 min threshold time. During rush hour (a), the accessibility of Seoul traffic conditions decreased, showing some non-accessible areas. Whereas during non-rush hour times, the green areas tended to cover the whole city (excluding the river and mountain areas).

In this study, a proximity analysis method experiment was conducted to derive a data-driven dynamic service area for mobile vehicles. This allowed for the evaluation of the service area for each vehicle within a 5 to 10 min threshold with a time-resolution of 30 min. Additionally, a pre-calculated road segment node dynamic service area was derived to quickly respond to service area queries.

The DTG data used in the study were collected from MOLIT; the data had been accumulating for over 1 year, and due to the large file size (800 GB per month on average), preprocessing was conducted to remove non-significant data. By implementing the data-driven service area analysis method, the dynamic service areas for vehicles within Seoul were derived with a 5 min temporal resolution. The time threshold of the service area was set as 5 min and 10 min, which could be altered by users to derive a user-defined time threshold service area. However, the 5 min threshold dynamic service area still contained some uncovered spaces, especially during rush hours.

Compared to conventional methods, the suggested dynamic service area analysis could consider traffic conditions, as shown in Figure 1a,b, since the method makes use of GPS data, which are also affected by real-world traffic conditions at any given time. The system could be implemented in both public and private sectors to draw new insights. In the emerging field of smart cities, when the system is implemented with police patrol car data, it could result in an optimized route for patrol, leading to a more efficient coverage area and quicker response time. If the system is implemented in private sectors, new insights that can benefit both suppliers and customers can be derived. For example, in the field of taxi call-dispatch systems, customers can be connected to taxis within an accurate 5 min service area. Additionally, when implemented in upcoming autonomous vehicles, the system could present various possibilities for auto-piloting cars to ensure maximum coverage and quick response times.