Research Themes|Mobility analysis / Modeling

Cooperative Vehicle Awareness using Vehicle-to-Vehicle Communication

It has become more common to disseminate traffic information and safe driving assistance information by utilizing Vehicle-to-Roadside (V2R) communications.Alternatively, cooperative awareness techniques by utilizing Vehicle-to-Vehicle (V2V) communications attract a lot of attention. In these techniques, each vehicle periodically sends its own position information via V2V and recognizes the existence of neighboring vehicles. However, in the early stage of penetration of on-board devices, it is not possible for a vehicle to acquire the information about vehicles which do not have communication devices. In contrast, with vehicle detection devices that aim at autonomous safety such as ranging sensors and vehicle on-board mounted cameras, more neighboring vehicles may be detected by those vehicles.

Based on these situations, I study a cooperative awareness method in which multiple vehicles cooperatively calibrate their positions and recognize neighboring vehicles to obtain accurate positions. In the proposed method, only limited vehicles are equipped with GPS receivers, DSRC devices and vehicle detection devices such as all directional sensors and cameras and share their own positions from GPS signals and relative positions calculated by observation among other equipped vehicles. Then, the proposed method aims that each equipped vehicle recognizes its neighboring vehicles including indirect neighbors by combining various pieces of position information.

From the result of simulation experiment, about 60% of neighboring vehicles which are located within a 500m radius of own position are recognized within 2.0m errors in the environment that the penetration rate is 30%.

Estimating Traffic Velocity Lowering in Snowy Season

In the snow city where a large amount of snow falls in the winter season,the snow has a big influence on the traffic flow.The traffic capacity decreases rapidly,because the snow on the road disturbs traveling of the cars,and the snow which was piled up on the roadside reduces the width of a road.It is important problem for traffic planning in the snow city to grasp the influence of the snow on road traffic.In this study, we collect weather data such as snowfall or snow amount and vehicle information provided from probe car data which is one of the traffic data collected in the world eminent snow city, Sapporo-shi.Then we suggest a method of making the model type that estimates traffic velocity lowering by analyzing these data by a multiple regression analysis.We evaluate degree of the influence of a change of the climatic condition on a drop of the traffic speed of the snow-covered road by making the model type on every route and analyzing it.We showed that traffic velocity lowering of the snow-covered road can explain by weather data to some extent,as a result of model construction by the suggested method using the probe car data on the real road in Sapporo-shi for plural routes.In addition, I considered the difference in speed drop factor every route.

Laser Range Scanner

Pedestrian tracking techniques have been of significant importance in recent years, enabling for a variety of human-centric applications like personal / crowd navigation, facility design, evacuation planning and path analysis. Although previous literature on pedestrian tracking techniques has shown that these systems can accurately capture pedestrian flows and their spatial distribution under specific experimental setups, I have found through a long-term field experiment in a public commercial building that there still remains big challenge to make them robustly work in real world: The system often misses some of the pedestrians in the monitored region, especially when the area of interest is crowded with many visitors. It may seriously harm effectiveness of the tracking system since presence and behavior of pedestrian crowds are essential information in estimating popularity of booths in an exhibition hall, building an efficient disaster evacuation strategies, and so on. In this paper, we propose and implement a novel system for simultaneously tracking pedestrians and human crowds by using a small number of laser range scanners (LRSs). In the areas, where crowd density is moderate, the proposed system accurately tracks individual persons’ trajectories to enable fine-grained path analysis. Meanwhile, for the extremely crowded regions, in which trajectories of individual persons can be hardly tracked in a continuous manner, it automatically switches its algorithm for accurately estimating the number of pedestrians in the area. For that purpose, I extract two feature values from the measurement data of LRSs, and build an empirical regression model to estimate crowd density under severe occlusion.

I have conducted simulations, a laboratory-scale experiment, and a long-term field experiment in a real exhibition space in a large commercial complex in the heart of Osaka downtown. From those results, I have confirmed that the algorithm and system can achieve practical and reasonable performance in a variety of situations.

Copyright(c) 2018 Higashino Lab. All Rights Reserved. Design by