PURPOSES : The reliability of traffic volume estimates based on location intelligence data (LID) is evaluated using various statistical techniques. There are several methods for determining statistical significance or relationships between different database sets. We propose a method that best represents the statistical difference between actual LID-based traffic volume estimates and the VDS values (i.e., true values) for the same road segment. METHODS : A total of 2,496 datasets aggregated for 1-h LID and VDS data were subjected to various statistical analyses to evaluate the consistency of the two datasets. The VDS data were defined as the true values for comparison. Four different statistical techniques (procrutes, 2-sample t-test, paired-sample t-test, and model performance rating scale) were applied. RESULTS : In cases where there is a specific pattern (e.g., traffic volume distribution considering peak and off-peak times), distribution tests such as Procrustes or Kolmogorov-Smirnov are useful because not only the prediction accuracy but also the similarity of the data distribution shape is important. CONCLUSIONS : The findings of this study provide important insight into the reliability of LID-based traffic volume estimation. To evaluate the reliability between the two groups, a paired-sample t-test was considered more appropriate than the performance evaluation measure of the machine-learning model. However, it is important to set the acceptance criteria necessary to statistically determine whether the difference between the two groups in the paired-sample t-test varies according to the given problem.

도시의 확장 및 광역화로 지상에서의 물리적인 도로 공급 및 확대는 현실적으로 어려운 상황에 직면하였으며, 이를 해결하기 위한 정책의 일환으로 경부고속도로 지하화 사업으로 대표되는 고속도로 지하화 사업이 추진되고 있다. 지하공간을 이용한 도로의 입체적 확장은 새로운 도로 공급용지의 공간적 확대뿐 아니라 지상도로의 교통량 분산으로 교통정체 완화, 차량으로 인한 소음 및 대기오염 문제 완화 등 도로교통체계의 효율성 및 문제점을 개선할 것으로 판단된다. 그러나 현재까지는 지하도로의 설계 및 시공 관련 기술을 위주로 개발 및 적용되고 있어, 운영 및 안전관리에 필수적인 지하고속도로 교통류 관리에 관한 연구는 이론적 수준에 머물고 있는 한계가 상존한 상황에 있다. 이에 본 연구는 지하고속도로 안전 향상을 위한 사고 예방 및 대응 기술 개발의 전단계로 유고 상황에 대응하기 위한 교통관리 개념을 유고 상황에 따라 변화되는 혼잡 지속시간을 추정할 수 있는 교통류 진단과 지하고속도로 내 유고 발 생 후 교통류 혼잡 회복 및 상태 안전화로 나누어 제안하였다. 이를 기반으로 향후 추진될 기술 개발의 단계를 교통류 변동 지표 개 발, 교통류 진단 알고리즘 개발, 교통류 해석ㆍ추정 알고리즘 개발, 교통관리 현장 적용 및 검증의 4단계로 기술 개발 로드맵을 제시 하였다. 본 연구의 결과는 향후 추진되는 K-지하고속도로 안전 확보를 위한 사고 예방 및 대응 기술 개발에 대한 실효성 및 신뢰성을 높이는데 기여할 것으로 사료된다.

PURPOSES : We propose a framework to evaluate the reliability of integrating homogeneous or heterogeneous mobility data to produce the various data required for greenhouse gas emission estimation. METHODS : The mobility data used in the framework were collected at a fixed time from a specific point and were based on raster data. In general, the traffic volume for all traffic measurement points over 24 h can be considered raster data. In the future, the proposed framework can be applied to specific road points or road sections, depending on the presence or absence of raster data. RESULTS : The activity data required to calculate greenhouse gas emissions were derived from the mobility data analysis. With recent developments in information, communication, and artificial intelligence technologies, mobility data collected from different sources with the same collection purpose can be integrated to increase the reliability and accuracy of previously unknown or inaccurate information. CONCLUSIONS : This study will help assess the reliability of mobility data fusion as it is collected on the road, and will ultimately lead to more accurate estimates of greenhouse gas emissions.

PURPOSES : This study aimed to identify factors affecting the duration of traffic incidents in tunnel sections, as accidents in tunnels tend to cause more congestion than those on main roads. Survival analysis and a Cox proportional hazards model were used to analyze the determinants of incident clearance times. METHODS : Tunnel traffic accidents were categorized into tunnel access sections versus inner tunnel sections according to the point of occurrence. The factors affecting duration were compared between main road and tunnel locations. The Cox model was applied to quantify the effects of various factors on incident duration time by location. RESULTS : Key factors influencing mainline incident duration included collision type, driver behavior and gender, number of vehicles involved, number of accidents, and post-collision vehicle status. In tunnels, the primary factors identified were collision type, driver behavior, single vs multi-vehicle involvement, and vehicles stopping in the tunnel after collisions. Incidents lasted longest when vehicles stopped at tunnel entrances and exits. In addition, we hypothesize that incident duration in tunnels is longer than in main roads due to the reduced space for vehicle handling. CONCLUSIONS : These results can inform the development of future incident management strategies and congestion mitigation for tunnels and underpasses. The Cox model provided new insights into the determinants of incident duration times in constrained tunnel environments compared to open main roads.

PURPOSES : This study investigates the factors affecting extra-long tunnel accidents by integrating data on tunnel geometry, traffic flow, and traffic accidents and derives the underlying implications to mitigate the severity of accidents. METHODS : Two processes centered on three key data points (tunnel geometry, traffic flow, and traffic accidents) were used in this study. The first is to analyze the spatial characteristics of extra-long tunnel traffic accidents and categorize them from multiple perspectives. The other was to investigate the factors affecting extra-long tunnel traffic accidents using the equivalent property-damage-only (EPDO) of individual accidents and the aforementioned data as the dependent and independent variables, respectively, by employing an ordered logistic regression model. RESULTS : Gyeonggi-do, Gyeongsangnam-do, and Gangwon-do are three metropolitan municipalities that have a significant number of extra-long tunnel accidents; Busan and Seoul have the most extra-long tunnel accidents, accounting for 23.2% (422 accidents) and 18.6% (339 accidents) of the 1,821 accidents that occurred from 2007 to 2020, respectively. In addition, approximately 70% of extra-long tunnel traffic accidents occurred along tunnels with lengths of less than 2 km, and Seoul and Busan accounted for over 60% of the top 20 extra-long tunnels with accidents. Most importantly, the Hwangryeong (down) tunnel in Busan experienced the most extra-long tunnel traffic accidents, with 77 accidents occurring during the same period. As a result of the ordered logistic regression modeling with EPDO and multiple independent variables, the significant factors affecting the severity of extra-long tunnel traffic accidents were determined to be road type (freeway, local route, and metropolitan city road), traffic flow (speed), accident time (year, summer, weekend, and afternoon), accident type (rear end), traffic law violations (safe distance violation and center line violation), and offending vehicles (van, sedan, and truck). CONCLUSIONS : Based on these results, the following measures and implications for mitigating the severity of extra-long tunnel traffic accidents must be considered: upgrading the emergency response level of all road types to that of freeways and actively promoting techniques for regulating high-speed vehicles approaching and traversing within extra-long tunnels are necessary. In addition, the emergency response and preparation system should be reinforced, particularly when the damage from extra-long tunnel traffic accidents is more serious, such as during the summer, weekends, and afternoons. Finally, traffic law violations such as safe distance and centerline violations in extra-long tunnels should be prohibited.

PURPOSES : Because a driving simulator typically focuses on analyzing a driver’s driving behavior, it is difficult to analyze the effect on the overall traffic flow. In contrast, traffic simulation can analyze traffic flow, that is, the interaction between vehicles; however, it has limitations in describing a driver’s driving behavior. Therefore, a method for integrating the simulator and traffic simulation was proposed. Information that could be controlled through driving experiments was used, and only the lane-change distance was considered so that a more natural driving behavior could be described in the traffic flow. METHODS : The simulated connection method proposed in this study was implemented under the assumption of specific traffic conditions. The driver’s lane-changing behavior (lane-changing distance, deceleration, and steering wheel) due to the occurrence of road debris was collected through a driving study. The lane-change distance was input as a parameter for the traffic simulation. Driving behavior and safety were compared between the basic traffic simulation setting, in which the driver's driving behavior information was not reflected, and the situation in which the driving simulator and traffic simulation were integrated. RESULTS : The number of conflicts between the traffic simulation default settings (Case 1) and the situation in which the driving simulator and traffic simulation were integrated (Case 2) was determined and compared for each analysis. The analysis revealed that the number of conflicts varied based on the level of service and road alignment of the analysis section. In addition, a statistical analysis was performed to verify the differences between the scenarios. There was a significant difference in the number of conflicts based on the level of service and road alignment. When analyzing a traffic simulation, it is necessary to replicate the driving behavior of the actual driver. CONCLUSIONS : We proposed an integration plan between the driving simulator and traffic simulation. This information can be used as fundamental data for the advancement of simulation integration methods.

PURPOSES : In this study, the factors affecting the severity of traffic accidents in highway tunnel sections were analyzed. The main lines of the highway and tunnel sections were compared, and factors affecting the severity of accidents were derived for each tunnel section, such as the tunnel access zone and tunnel inner zone. METHODS : An ordered probit model (OPM) was employed to estimate the factors affecting accident severity. The accident grade, which indicates the severity of highway traffic accidents, was set as the dependent variable. In addition, human, environmental, road condition, accident, and tunnel factors were collected and set as independent variables of the model. Marginal effects were examined to analyze how the derived influential factors affected the severity of each accident. RESULTS : As a result of the OPM analysis, accident factors were found to be influential in increasing the seriousness of the accident in all sections. Environmental factors, road conditions, and accident factors were identified as the main influential factors in the tunnel access zone. In contrast, accident and tunnel factors in the tunnel inner zone were found to be the influencing factors. In particular, it was found that serious accidents (A, B) occurred in all sections when a rollover accident occurred. CONCLUSIONS : This study confirmed that the influencing factors and the probability of accident occurrence differed between the tunnel access zone and inner zone. Most importantly, when the vehicle was overturned after the accident occurred, the results of the influencing factors were different. Therefore, the results can be used as a reference for establishing safety management strategies for tunnels or underground roads.

PURPOSES : It is necessary to implement traffic-control strategies for underground roads. In this study, the application criteria for traffic control were developed to minimize actual traffic congestion on underground roads before it occurs. In particular, the traffic congestion judgement criteria and procedure (TJCAP) were developed. They can specifically classify the possibility of traffic congestion underground. METHODS : A microscopic traffic simulation model was used to analyze different scenarios. With the scenario simulation results, a hierarchical clustering analysis was applied to produce quantitative values from the TJCAP for each experimental network case. RESULTS : For network case (a), it was concluded that the possibility of traffic congestion on underground roads increases when the speed of the ground road connected to the main underground road and the connected ground road after the outflow of the ramp section is low. When the connected road is an interrupted facility after entering the underground roads, the red time is long, and when the section travel speed is 15 km/h, the possibility of traffic congestion underground is highest. A cluster analysis based on these results was performed using two techniques (elbow and silhouette) to verify the final classification. CONCLUSIONS : The TJCAP were designed to operate traffic flow with stricter criteria than traffic congestion management on ground roads. This reflects the difference in the driving environment between underground and above-ground roadways.

PURPOSES : In this study, model-agnostic methods are applied for interpreting machine learning models, such as the feature global effect, the importance of a feature, the joint effects of features, and explaining individual predictions. METHODS : Model-agnostic global interpretation techniques, such as partial dependence plot (PDP), accumulated local effect (ALE), feature interaction (H-statistics), and permutation feature importance, were applied to describe the average behavior of a machine learning model. Moreover, local model-agnostic interpretation methods, individual conditional expectation curves (ICE), local surrogate models (LIME), and Shapley values were used to explain individual predictions. RESULTS : As global interpretations, PDP and ALE-Plot demonstrated the relationship between a feature and the prediction of a machine learning model, where the feature interaction estimated whether one feature depended on the other feature, and the permutation feature importance measured the importance of a feature. For local interpretations, ICE exhibited how changing a feature changes the interested instance’s prediction, LIME explained the relationship between a feature and the instance’s prediction by replacing the machine model with a locally interpretable model, and Shapley values presented how to fairly contribute to the instance’s prediction among the features. CONCLUSIONS : Model-agnostic methods contribute to understanding the general relationship between features and a prediction or debut a model from the global and/or local perspective, securing the reliability of the learning model.

PURPOSES : This study prioritizes the potential technology for establishing an efficient traffic control in the ramp junction of urban deep underground tunnels in the future. We considered most of the applicable technologies that ensure traffic safety at the on-off ramp junction. METHODS : This study proposes a methodology to prioritize the applicable technology for establishing efficient traffic control in the ramp junction of an urban deep underground tunnel using an analytical hierarchy process (AHP). First, an AHP structure was developed. Second, an individual survey was conducted to collect the opinions of road and transportation experts. Based on the survey results, weights were estimated depending on the relevant criteria of the developed structure. The estimated weights were verified using the consistency index (CI) and consistency ratio (CR). In addition, a sensitivity analysis was performed to confirm the reliability of the estimated weights. Finally, the potential technology for an efficient traffic control in the ramp junction of an urban deep underground tunnel was prioritized. RESULTS : In the first level of hierarchy, traffic demand control had the highest priority, and ramp metering, section speed control, and shoulder lane control were selected in the second level of hierarchy. CONCLUSIONS : These results implied that prioritizing would be useful in establishing traffic operation strategies for traffic safety when constructing and opening deep underground tunnels in urban areas in the future.

PURPOSES : The purpose of this study is to derive specific road design elements for safe urban underground and to adopt measures for minimizing traffic delays and to maintain efficient operation. METHODS : In this study, a qualitative study was conducted using Focus Group Interview (FGI) method to identify significant connection characteristics and develop connections to urban underground roads. Finally, this study analyzes design elements necessary for traffic safety and efficient traffic operation. In addition, relevant case studies were performed with keywords from the FGI method results. Therefore, major design elements were analyzed for urban underground road connection and connection analysis for traffic simulation-based verification. RESULTS : The main characteristics of the connection between the underground roads of the downtown area were divided into three types: traffic flow characteristics, geometric characteristics, and driver behavioral characteristics. From the review of 16 leading studies (10 domestic papers and 6 international papers) according to the characteristics, the main design factors for “traffic flow characteristics” include the traffic volume, design speed, heavy vehicle ratio, and lane change. The important design elements for “geometric characteristics” include the separation distance, number of lanes, slope, lane and shoulder width and the design factors for “driver behavioral characteristics” showed reaction time, driver vision, and driving speed. CONCLUSIONS : The FGI method identified the main characteristics of connections to the underground roads. In addition, the relevant empirical and theoretical research data were considered in case studies, and the design elements were derived and separated spatially based on the features of each design element, establishing a point-specific design element guideline.

PURPOSES: This study presents the results of the collection rate of various road debris dummies using Automated Road Debris Remover System(ROBOS), which is a newly developed automatic road-debris removal system. In addition, for traffic flow safety, appropriate safe deceleration distances for ROBOS are estimated using VISSIM. METHODS: A total of 12 kinds of road debris dummies were selected based on the opinions of public agencies, and randomly placed on the road. Repeated tests were performed. The road debris dummies were placed on the center of the lane, the shoulders, and the median. During the test, the ROBOS running speed was maintained at 15 km/h for an approximately 10-km-long roadway, and the collection and loading process was performed five times under the same condition. For the simulations, the road debris was assumed to be placed on Lane 1 under different traffic conditions and grades. Multiple simulations were conducted, and the average values were used to obtain appropriate safe deceleration distances for ROBOS. RESULTS: The dummies were considerably large and heavy, and the collection rate was very high for the very light dummy. The simulation results indicate that during debris collection using ROBOS, when the traffic volume is relatively low, the degree of change in traffic flow is high, and when the traffic flow rate is high, the risk of an accident is also considerable. CONCLUSIONS : We demonstrated that the debris collection efficiency of ROBOS is very high when the brush-conveyor and vacuum are used simultaneously. Further, when the exact location of road debris on specific lanes and for specific traffic volumes is known, a safe deceleration distance is recommended.

PURPOSES: This study aimed to evaluate the performance of a model developed for road surface temperature change pattern in reflecting specific road characteristics. Three types of road sections were considered, namely, basic, tunnel, and soundproof tunnel. METHODS: A thermal mapping system was employed to collect actual road surface temperature and locational data of the survey vehicle. Data collection was conducted 12 times from 05:30 am to 06:30 am on the test route, which is an uninterrupted flow facility. A total of 9010 road surface temperature data were collected, and half of these were selected based on a random selection process. The other half was used to evaluate the performance of the model. The model used herein is based on machine learning algorithms. The mean absolute error (MAE) was used to evaluate the accuracy of the estimation performance of the model. RESULTS: The MAE was calculated to determine the difference between the estimated and the actual road surface temperature. A MAE of 0.48℃ was generated for the overall test route. The basic section obtained the smallest error whereas that of the tunnel was relatively high. CONCLUSIONS: The road surface temperature change is closely related to the air temperature. The process of data pre-processing is very important to improve the estimation accuracy of the model. Lastly, it was difficult to determine the influence of the data collection date on the estimation of the road surface temperature change pattern due to the same weather conditions.

PURPOSES: This study develops various models that can estimate the pattern of road surface temperature changes using machine learning methods. METHODS : Both a thermal mapping system and weather forecast information were employed in order to collect data for developing the models. In previous studies, the authors defined road surface temperature data as a response, while vehicular ambient temperature, air temperature, and humidity were considered as predictors. In this research, two additional factors-road type and weather forecasts-were considered for the estimation of the road surface temperature change pattern. Finally, a total of six models for estimating the pattern of road surface temperature changes were developed using the MATLAB program, which provides the classification learner as a machine learning tool. RESULTS: Model 5 was considered the most superior owing to its high accuracy. It was seen that the accuracy of the model could increase when weather forecasts (e.g., Sky Status) were applied. A comparison between Models 4 and 5 showed that the influence of humidity on road surface temperature changes is negligible. CONCLUSIONS: Even though Models 4, 5, and 6 demonstrated the same performance in terms of average absolute error (AAE), Model 5 can be considered the optimal one from the point of view of accuracy.