Scanning the Issue

Azim Eskandarian
2020 IEEE transactions on intelligent transportation systems (Print)  
An analysis of more than five decades of research on pedestrian behavior understanding followed by a review of autonomous driving technologies designed for interaction with pedestrians is presented. The meta-analysis of past literature identifies more than 30 factors, ranging from social to environmental conditions, which can potentially influence the way pedestrians behave. Based on these findings and the current state-of-the-art solutions for interaction between autonomous vehicles (AVs) and
more » ... edestrians, a road map for future research is proposed regarding what methods and modalities of communication should be used by AVs, what information should be considered by algorithmic solutions for accurate prediction of pedestrians' intentions, and what behavioral studies are required to better understand pedestrians' actions in various situations. A coevolutionary algorithm for two solution sets, population and archive, using objective-wise local search and set-based simulated binary operation in order to address the multiobjective carpool service problem with time windows. In the evolution module of the proposed algorithm, three different methods, namely, objective-wise local search in an archive, set-based simulated binary operation in a population, and setbased simulated binary operation both in a population and in an archive, were used to generate the offspring. The results of the quantitative comparison and objective visualization showed that the proposed algorithm can obtain superior Pareto-optimal solutions regarding convergence and diversity compared with a fast nondominated sorting genetic algorithm. The pantograph is one of the most important components in electrical railway vehicles. Periodical inspection and maintenance of the pantograph slide plate are significant in terms of safe and stable operation. In this article, an innovative and intelligent method based on deep learning and image processing technologies is proposed for online condition monitoring of the pantograph slide plate. In the first part, the surface defect detection and recognition method of the pantograph slide plate is proposed. Four typical surface defects of the Digital Object Identifier 10.1109/TITS.2020.2972652 slide are considered and a deep learning model, PDDNet is trained for defect detection and recognition. In the second part, the wear edge estimation based on image processing technology is investigated in detail. Furthermore, they are used to calculate the wear depth and evaluate the wear condition of the pantograph slide. The wear depth estimation results are compared with on-site measurement data. This work excogitates a DCNN model for video foreground/background segmentation. Initially, a Conv-LSTM2D image to image encoder-decoder (EnDec) model is designed. Then, using it as a baseline, a 3D CNN-LSTM EnDec network is implemented. The new architecture is optimized for trainable parameters but more in-depth than the baseline as it contains micro-autoencoders and slow decoding blocks with frequent residual feature forwarding. It captures short-and long-term spatiotemporal cues through 3D convolutions and LSTM units from a set of t frames before predicting the segmentation of the current frame. The model is a new addition to the state-of-the-art FG-BG segmentation algorithms as it produces very competitive performances on diverse conditions, like lighting variations, cast shadow, dynamic backgrounds, and nighttime in indoor and outdoor environments. The results show that the model superiorly performs when compared with traditional and modern NN-based methods. However, it lacks the capability of handling moving camera scenarios. Urban traffic passenger flows prediction is practically important to facilitate many real applications including transportation management and public safety. In this article, the authors propose a deep irregular convolutional residual LSTM network model called DST-ICRL for urban traffic passenger flow prediction. They first modelled the passenger flows among different traffic lines in a transportation network into multichannel matrices analogous to the RGB pixel matrices of an image. Then, they propose a deep learning framework that integrates irregular convolutional residential network and LSTM units to learn the spatial-temporal feature representations. Third, they fuse other external factors further to facilitate a real-time prediction. They conduct extensive experiments on different types of traffic passenger flow data sets including subway, taxi, and bus flows in Beijing as well as bike flows in New York City. The results show that the proposed DST-ICRL 1524-9050
doi:10.1109/tits.2020.2972652 fatcat:ppev3ra7hbeopio2yqbbf32gbu