Earticle

Drowsy driving of commercialized vehicles has been a challenging issue to road safety hazards. Among the vehicles, hazardous materials transport trucks should be addressed to prevent not only critical road hazards but also secondary disasters by their drowsy driving accidents. Here, real-world driving data and driving environmental data were utilized to evaluate fatigue condition intensity (FCI) on highways about drowsy driving to hazardous material vehicles. With the Drowsy driving dataset constructed based on drowsy driving features in previous studies, the study proposed a methodology to compute FCI to fatigue conditions by rule-based association analyses. The proposed methodology offers numerical indicators of FCI of the drivers and presents similar insights to fatigue conditions with other research. In addition, the numerical indicator contributes to describing microscopic features of drowsy driving conditions.

This paper presents the data-driven modeling based on recurrent neural network (RNN) for the system identification of vehicle dynamics. For the accurate control of an autonomous vehicle, a sophisticated model of the vehicle dynamics is required. Data-driven modeling can achieve such a model only with the data obtained from the target. A simulation result by using the data from a driving simulator is included to test the feasibility of the RNN for the system identification of a ground vehicle.

This study has implemented before and after analysis for traffic accident rate on the 38 IC sections of 9 expressways during three years before construction, construction periods, and three years after construction respectively. The number of accident and fatalities had slightly increased during construction period compared to before construction and then had decreased after the construction was completed. In the case of the number of injured, it had continuously decreased through three periods. Especially, it had sharply dropped after the construction was completed. According to the paired t-test, the average of accident between the construction period and the period after construction is different. The average of fatalities is different between before and after the construction, and also during and after the construction. The average of injured is totally different at every three periods. The number of accident, injured, and fatalities had largely decreased by 41, 32, and 62 percent respectively when is compared to before and after expressway widening project. It could support the hypothesis that the widening project could result in lower accident rate while increasing road capacity. This study has a value as a comprehensive before and after study for traffic accident rate on the widening project section of expressway, which is hard to find out similar studies in the past due to data limitation. The result of this study could use as a basic unit (per unit) of traffic accident rate when the preliminary feasibility study is implemented for the widening project. The current method of estimating traffic accident cost savings has a significant limitation in reflection of impact on safety improvement because it is calculated based on vehicle-km. Therefore, traffic congestion level needs to be considered in the process of preliminary feasibility study for the highway project.

This paper presents an interim result of a study on virtual LiDAR sensor modeling for autonomous driving simulators. Based on a basic knowledge on the characteristics of pulsed light beams, which may be dependent on the color and material of the object to be reflected, a simple model of equation for the intensity is used for modeling the reflected power of beams of LiDAR. Experiments using a real LiDAR and a virtual one to distinguish colors reveals that the accessibility to colors of the objects is essential for modeling of LiDAR sensors in autonomous driving simulators.

Understanding an accurate trip demand by purpose is crucial for short-term regional planning but also long-term regional planning. In traditional approach, information of purpose-oriented trip demand has been derived from public survey in South Korea such as Travel Diary Survey. This type of data acquiring method maybe useful in a sense that it can capture a meaningful sample regarding to entire country, meanwhile it costs a tremendous amount of budget and time. In this research, we want to offer a novel framework for estimating purposeoriented trip demand with dynamic and effective fashions using data fusion in conjunction with Machine learning techniques. With primary results of this concept, this study showed how several state-of-the-art algorithms, including Deep neural network, UMAP, and random forest in conjunction with Genetic algorithm and Tabu-search for optimization, can contribute to this framework. Although tangible results are yet to come, we expect this framework can contribute to resilience planning such as COVID-19.

Since the popularity of city monitoring systems has increased, surveillance videos have become more common. However, due to the massive increment of social event needs after lifting the coronavirus pandemic lockdown, the pressure of handling anomalies on government departments is growing with each passing day. In addition, surveillance cameras cannot record every corner of our city. There are still blind areas such as indoor environments and remote regions. Therefore, this paper introduces and develops an comprehensive system for video anomaly detection in smart city management, which is based on mobile phones, to enlarge the coverage of the current surveillance system. This system can capture and transmit abnormal videos. In addition, a deep learning method (VGG16) is implemented in the system to do video anomaly detection.