The epidemic of the new coronavirus infection has had a great impact on all fields of society, but the impact is obvious in the field of mobility, and measures such as a significant decrease in traffic demand and implementation of infectious disease countermeasures in transportation facilities was actively carried out. On the other hand, it is expected that some of the changes that have occurred in people's lives and social structures, such as online substitution of various activities, will remain irreversible even after an infectious disease epidemic. In this paper, we will discuss the ideal future of mobility and the direction to aim for in light of the changes of the times.
Although automated driving (AD) technology has been considered to be a good match for sharing services, user's psychological resistance to use various types of the sharing services by using the AD technology is expected to increase in a society that has experienced COVID-19 pandemic. The amount of human movement decreased, and B-to-C logistics increased significantly, as a result of the requirement for behavioral change to avoid enclosure, congestion, and closeness as a countermeasure against infectious diseases. This manuscript discusses not only the situations where the AD technology is required and effective but also the future prospects of the AD technology such as what new technologies are expected to be developed in the future.
Drivers’ behaviors management is one of the most important aspects of signalized intersection design. Signal head placement causes differences at the start-up as well as stopping and crossing behaviors from perspectives of safety, ergonomics and performance. The research discusses visibility conditions and blockage of signal heads by leading vehicles based on practices in USA, Japan, Russia, UK, France and Germany. As a result, it was found that near side signal head placement at the vicinity of stop line may result in lower performance in vehicles’ start-up due to worse signal visibility. In contrast, far side strategy sometimes generates unsafe conditions in signal blockage scenario.
Reinforcement Learning (RL) methods have been introduced to traffic control application for several decades. Traditional RL-based signal controls are mostly model-free that ignore the complex traffic states variation. Such treatment is not realistic due to external uncertainty of traffic. To fill this gap, an independent prediction module could be introduced to formulate a model-based RL. This study introduces queuing estimation models into deep-Qnetwork(DQN)-based signal control. The queuing situation could be reproduced and predicted by both input-output model and shock wave model. Through the empirical experiment, we confirm the necessity of prediction in RL-based signal control for isolated intersection.
This study addresses the design of exits for dedicated Connected-and-Automated-Vehicle (CAV) lanes. Because vehicles from such lanes have to merge with the adjacent normal lane with Human-Driven-Vehicles (HDVs), the exit locations must be carefully determined according to the HDV gap distribution. Therefore, this study aims to understand merging potential by analyzing HDV gap distributions along motorway sections. For this, the “available percentage” is defined as the percentage of gaps longer than the assumed critical gap. By analyzing the data collected on the Hanshin Expressway in Japan, we found that the available percentage differs by road geometry, especially during free-flow conditions.
Connected and automated vehicles (CAVs) may drive on dedicated lanes in the near future. For better safety and efficiency, lane changings between dedicated lanes and regular lanes are only allowed in certain areas, and the lane changing maneuvers in such areas are complicated and worthy of investigation. In this paper, we propose a dynamic trajectory planning methodology for CAVs to complete lane changings in the restricted allowing areas. The trajectory planning process utilizes real-time status of other relevant vehicles, and involves a tradeoff among safety, comfort and efficiency. A numerical simulation is conducted to verify the performance of the proposed methodology.
In Japan, the importance of accumulating urban and/or daily-life facilities within compact areas (called “centers”) and connecting these centers using a transportation network, known as the “Compact + Network” concept, has been recognized. Accordingly, road network evaluation should be performed by considering the relationship of the network with urban center locations. However, the size and shape of service spheres that an urban center should cover differ depending on various factors, such as terrain and population distribution. In this article, such differences were visualized using wind rose, based on the bearing and distance from settlements to their nearest urban center. The features that should be addressed when evaluating road network topology and road travel performance were discussed.
Data mining based vehicle probe data analysis has addressed many attentions since we have entered the era of big data. In this work, we propose a study of combining probe data with weather information to perform sudden braking estimation using data mining techniques. We train a deep neural network and estimate weather situations from drive recorders in real time. In addition, we also gather information from meteorological observatories to provide further comparisons. Our experimental results show that the usage of weather information have a slight performance improvement over probe data only analysis.
Acoustic Vehicle Alerting System (AVAS) alerts pedestrians to the approach of the electric vehicle with a low noise level when traveling at a low speed. This system may annoy pedestrians depending on the surrounding environment and conditions. Therefore, a system that is highly acceptable to pedestrians while maintaining safety is required. In this study, the AVAS that uses deep learning to determine pedestrians who need to present information and appropriately presents information to those pedestrians by directional speakers was devised. Then, the safety and acceptability of the system were evaluated by actual vehicle experiments. The experimental results show that this AVAS tends to provide information only to appropriate pedestrians while maintaining safety.