Region-level traffic information can characterize dynamic changes of urban traffic at the macro level. Real-time region-level traffic prediction help city traffic managers with traffic demand analysis, traffic congestion control, and other activities, and it has become a research hotspot. As more vehicles are equipped with GPS devices, remote sensing data can be collected and used to conduct data-driven region-level-based traffic prediction. However, due to dynamism and randomness of urban traffic and the complexity of urban road networks, the study of such issues faces many challenges. This paper proposes a new deep learning model named TmS-GCN to predict region-level traffic information, which is composed of Graph Convolutional Network (GCN) and Gated Recurrent Unit (GRU). The GCN part captures spatial dependence among regions, while the GRU part captures the dynamic change of traffic within the region. Model verification and comparison are carried out using real taxi GPS data from Shenzhen. The experimental results show that the proposed model outperforms both the classic time series prediction model and the deep learning model at different scales.

Transport-related problems, such as automobile dependence, traffic congestion, and greenhouse emissions, lead to a great burden on the environment. In developing countries like China, in order to improve the air quality, promoting sustainable travel modes to reduce the automobile usage is gradually recognized as an emerging national concern. Though there are many studies related to the physically active modes (e.g., walking and cycling), the research on the influence of attitudes to active modes on travel behavior is limited, especially in China. To fill up this gap, this paper focuses on examining the impact of attitudes to walking and cycling on commute mode choice. Using the survey data collected in China cities, an integrated discrete choice model and the structural equation model are proposed. By applying the hybrid choice model, not only the role of the latent attitude played in travel mode choice, but also the indirect effects of social factors on travel mode choice are obtained. The comparison indicates that the hybrid choice model outperforms the traditional model. This study is expected to provide a better understanding for urban planners on the influential factors of green travel modes.

Facing the potential dangers from sudden disasters in urban cities, emergency administrators have to make an appropriate evacuation plan to mitigate negative consequences. However, little attention has been paid to evacuee real decision psychology when developing a strategy. The aim of this paper is to analyze evacuee mode choice behavior considering regret aversion psychology during evacuation. First, the utility-based and regret-based models are formulated to obtain evacuees’ preferences on travel mode choice, respectively. According to the data collected from the stated preference (SP) survey on evacuee mode choice, the estimation results show that the regret-based model performs better than the utility model. Moreover, based on the estimates from behavioral analysis, the elasticities of evacuee mode choices are calculated, and transit strategy simulation is undertaken to investigate the influence on evacuee mode switching from private automobile to public transit. The results are expected to help emergency administrators to make a transit-oriented strategy for a sustainable evacuation plan, especially for the benefit of carless people.

Decision-making is an important component of autonomous driving perception, decision-making, planning, and control pipeline, which undertakes the task of how the ego vehicle makes high-level decision-making behaviors (such as lane change and car following) after sensing the environmental state, and then these high-level decision-making behaviors can be transmitted to the downstream planning and control module for specific low-level action execution. Based on the method of deep reinforcement learning (specifically, Deep Q network (DQN) and its variants), an integrated lateral and longitudinal decision-making model for autonomous driving is proposed in a multilane highway environment with both autonomous driving vehicle (ADV) and manual driving vehicle (MDV). The classic MOBIL and IDM models are used for the lateral and longitudinal decisions of MDV (i.e., lane changing and car following), while the lateral and longitudinal decisions of ADV are dominated by deep reinforcement learning models. In addition, this paper also uses the nonlinear kinematic bicycle model and two-point visual control model to realize the low-level control of both MDV and ADV. By setting a reasonable state, action, and reward function, this paper has carried out a large number of simulation experiments on the proposed autonomous driving decision-making model based on deep reinforcement learning in a three-lane road environment. The results show that under such scenario setting conditions, the deep reinforcement learning-based model proposed in this paper performs well in autonomous driving safety and travel efficiency. At the same time, when compared with the classical rule-based decision-making model (MOBIL&IDM), it is found that the model proposed in this paper can significantly achieve better results in episode rewards after stable training. In addition, through a large number of hyper-parameter tuning experiments, the performance of DQN, DDQN, and dueling DQN models, which are also deep reinforcement learning-based decision-making models, under different hyper-parametric configurations is compared and analyzed, which can provide a valuable reference for the specific scenario application of these models.

This research proposed an eco-approach and departure system for left-turn vehicles at a fixed-time signalized intersection. This system gives higher priority to enhancing traffic safety than improving mobility and fuel efficiency, and optimizes the entire traffic consisted of connected and automated vehicles (CAVs) and conventional human-driven vehicles by providing ecological speed trajectories for left-turn CAVs. All the ecological speed trajectories are offline optimized before the implementation of system. The speed trajectory optimization is constructed in Pontryagin’s Minimum Principle structure. The before and after evaluation of the proposed system shows the percentage of vehicles that drive pass the intersection at safe speed increases by 2.14% to 45.65%, fuel consumption benefits range 0.53% to 18.44%, emission benefits range from 0.57% to 15.69%, no significant throughput benefits is observed. The proposed system significantly enhances the traffic safety and improves the fuel efficiency and emission reduction of left-turn vehicles with no adverse effect on mobility, and has a good robustness against the randomness of traffic. The investigation also indicates that the computation time of proposed system is greatly reduced compared to previous eco-driving system with online speed optimization. The computation time is up to 0.01 s. The proposed system is ready for real-time application.