ABSTRACT: This study proposes a method to calculate driving behavior indices that assess collision risk of right turns at intersections using an inexpensive system (e.g., a car navigation system). This method considers characteristics of common vehicle trajectories to complement the disadvantage of an inexpensive system: low accuracy of vehicle position. We analyzed a right-turn dataset in the real world to identify the characteristics and proposed methods that correct vehicle position and calculate the indices using the correction. Validation results showed that the proposed method reduces errors and increases the correlation between the calculated and true values compared to the original method.

View full abstract

ABSTRACT: This article provides a systematic review of research articles on pedestrians and cyclists’ intention recognition to be integrated into autonomous vehicles, especially for decision making and motion planning. We firstly describe why the intention recognition of pedestrians and cyclists is suitable and necessary for autonomous vehicles and why they cannot only rely on traffic regulation laws. Then, we summarise, amongst others, the methodology and sensors used by eighteen peerreviewed research articles published in relevant conferences and journals. We performed a systematic review of articles of the last 10 years from the following databases: IEEE Xplore, Science Direct, ACM digital library, Springer Link, MDPI and Web of Science. We observe from the collected articles that most of them are relying on several sensors, with a predominance including video. They mostly try to obtain the probability of crossing or the trajectory of the pedestrian/cyclist, mostly using a Recurrent Neural Network. In addition to their algorithmic contribution, 4 studies also provide a dataset. We conclude this article by talking about the remaining open challenges.

View full abstract

ABSTRACT: Lane line is one of the important environmental information used in various self-localization methods. However, the effect of lane lines fading or wears overtime on localization accuracy is not clear. This paper investigates effects of road markings wear on ZNCC-based map matching accuracy. As a priori or an observation, LiDAR-based intensity maps of the surrounding environment within a certain height range from the road surface were generated. The true position of ego-vehicle is obtained using the post-processed data of GNSS-IMU measurements. In map matching the Zero-means Normalized Cross- Correlation (ZNCC) between maps of prior-knowledge and observation was used. The pixel of the maximum correlation is regarded as the matched position. The results indicated that frequent map updates can reduce lateral errors to less than ±0.25 m regardless of the actual state of the white line. Insufficient map updates can cause negative effects on matching at road markings. For robustness against changes in the surrounding environment, it is ideal to increase the contribution degree of lane lines by repairing lane lines and updating the map at simultaneously.

View full abstract

ABSTRACT: A membrane filter was fabricated using gas dispersant micro-sized Three-way Catalyst (TWC) particles on the conventional filter substrate. The porosity of the cross-sectional area of the TWC-membrane was measured through electron microscopy image processing technique. In the membrane manufacturing process, a parametric study was conducted by comparing three mean particle sizes under three superficial velocities. According to high resolution electron micrographs, the cross-sectional view of a single TWC-particle was revealed as a homogeneously agglomerated particle. The porosity of the cross-sectional area of the membrane was measured as approximately 64.4% with a slight difference about 3% depending on varying mean particle sizes and superficial velocities.

View full abstract