PURPOSES : This study develops a model that can estimate travel speed of each movement flow using deep-learning-based probe vehicles at urban intersections. METHODS : Current technologies cannot determine average travel speeds for all vehicles passing through a specific real-world area under obseravation. A virtual simulation environment was established to collect information on all vehicles. A model estimate turning speeds was developed by deep learning using probe vehicles sampled during information processing time. The speed estimation model was divided into straight and left-turn models, developed as fully-offset, non-offset, and integrated models. RESULTS : For fully-offset models, speed estimation for both straight and left-turn models achieved MAPE within 10%. For non-offset models, straight models using data drawn from four or more probe vehicles achieved a MAPE of less than 15%. The MAPE for left turns was approximately 20%. CONCLUSIONS : Using probe-vehicle data(PVD), a deep learning model was developed to estimate speeds each movement flow. This, confirmed the viability of real-time signal control information processing using a small number of probe vehicles.

PURPOSES : This study aims to perform a quantitative analysis of Forward Collision Warning and crash frequency using heavy vehicle driving data collected in expressway driving environments, and to classify the driving environments where Forward Collision Warnings of heavy vehicles occur for accident-prone areas and analyze their occurrence characteristics. METHODS : A bivariate Gaussian mixture model based on inter-vehicle distance gap and speed-acceleration parameters is used to classify the environment in which Forward Collision Warning occurs for heavy vehicles driving on expressways. For this analysis, Probe Vehicle Data of 80 large trucks collected by C-ITS devices of Korea Expressway Corporation from May to June 2022. Combined with accident information from the past five years, a detailed analysis of the classified driving environments is conducted. RESULTS : The results of the clustering analysis categorizes Forward Collision Warning environments into three groups: Group I (highdensity, high-speed), Group II (high-density, low-speed), and Group III (low-density, high-speed). It reveals a positive correlation between Forward Collision Warning frequency and accident rates at these points, with Group I prevailing. Road characteristics at sites with different accident incidences showed that on-ramps and toll gates had high occurrences of both accidents and warnings. Furthermore, acceleration deviation at high-accident sites was significant across all groups, with variable speed deviations noted for each warning group. CONCLUSIONS : The Forward Collision Warning of heavy vehicles on expressways is classified into three types depending on the driving environment, and the results of these environmental classifications can be used as a basis for building a road environment that reduces the risk of crashes for heavy vehicles.

교통사고 원인분석 및 사고예방을 위해서는 교통사고 유발요인에 대한 이해가 필요하다. 기존 연구에서는 기하구조, 운전자 특성 등의 요인을 고려하여 연구를 진행하였다. 그러나, 운전자 특성요인 분석에 사용된 자료는 검지기에서 측정된 집계된 속도로써, 속도 변화량을 이용한 사고분석연구에는 한계가 존재한다. 따라서, 본 연구에서는 차량의 속도변화 등의 수집이 용이한 센서를 이용하여 자료를 수집하였다. 가속도자료 및 기하구조 특성을 나타내는 변수를 설정하고, 사고자료와 매칭을 통해 사고개연성이 높은 잠재적 변수로의 적합성을 평가하였다. T-test, 이항 로지스틱 회귀분석을 사용했으며, T-test 결과로써 도출된 변수를 이항 로지스틱 회귀분석의 독립변수에 적용하고, 사고발생 유 무를 종속변수로 설정하였다. 분석결과, 5개의 변수가 사고발생에 영향을 주는 변수로 도출되었다. 또한, 도출된 모형은 사고발생구간의 예측에 적용할 수 있는 타당성을 확보하는 것으로 분석되었다. 본 연구에서 도출된 위험 운전행태 변수 및 모형은 프로브차량에 설치하여 활용할 수 있는 장치 등에 적용시켜 사고위험도 및 안전성 평가에 활용할 수 있을 것으로 기대된다.