24 April 2024 in IEEE Transactions on Intelligent Transportation Systems
Recently, the limited capacity of the Urban Rail Transit (URT) has failed to meet passenger travel demands, especially in peak hours, which leads to crowded stations and oversaturated metro networks. Considering the diverse scenarios of the multimodal public transportation network and the need for rapid generation of strategies, this study proposes a deep reinforcement learning (RL) algorithm that guides passenger route selection in the multimodal transit network to operate better. The proposed RL algorithm, consisting of graph representation learning, convolution neural network, and self-attention mechanism, is used to generate the route guidance strategy for passengers, aiming to alleviate the congestion of the multimodal transit network, improve passengers’ travel experience, and reduce CO $_{2}$ emissions. Based on the multimodal transit network in Beijing, the simulation results demonstrate that the RL algorithm can well perceive the states and generate adaptive route guidance strategies that can decrease the section load rates, improve network-wide passengers’ travel experience, and reduce CO $_{2}$ emissions. Even in cases where passengers are not fully compliant with the route guidance, the proposed algorithm remains effective.
23 May 2024 in Transportation Research Record: Journal of the Transportation Research Board
Urban rail transit (URT) has emerged as a crucial mode of transportation in metropolitan areas. For the effective operation of expanding URT networks, accurate short-term origin–destination (OD) demand distribution predictions are essential. This study introduces a novel deep-learning-based model for predicting short-term OD distribution in extensive networks, taking destination choice behaviors into account. First, we perform a comprehensive analysis of station passenger flows and OD flows from both temporal and spatial dimensions. Then, we develop the origin–destination distribution prediction (ODDP) model, combining the destination choice model (DCM) with the deep learning model (DLM). The DCM aims to understand OD distribution patterns from a behavioral perspective by transforming real-time inflows into OD distributions. Meanwhile, the DLM, employing attention and convolution layers, effectively captures the intricate temporal and spatial dynamics of passenger flows. Our model is evaluated using data from the Guangzhou Metro network in China, showing significant enhancements in prediction accuracy, model interpretability, and overall robustness. The implementation of our model promises substantial benefits for the operational efficiency of URT systems.
23 April 2026 in IEEE Transactions on Intelligent Transportation Systems
Pedestrian-vehicle interactions are critical problems in the development of autonomous driving technologies. However, there is still a lack of comprehensive understanding of pedestrian behaviors during pedestrian-vehicle interactions, especially considering the impact of group behaviors. This study proposes an Actor-Relation-Graph-based Game-Adversarial Inverse Reinforcement Learning (AR2G-AIRL) algorithm to understand pedestrian behaviors during pedestrian-vehicle interactions, considering the influence of surrounding pedestrians and vehicles. Specifically, we first identify pedestrian-vehicle interactions from a naturalistic driving dataset as the data basis. Subsequently, the AR2G-AIRL algorithm is constructed with three components: Actor-Relation Graphs (ARGs) to capture the features within groups, evolutionary game theory to extract the features between groups, and the Adversarial Inverse Reinforcement Learning (AIRL) algorithm to recover the reward function. The experiment results illustrate that the AR2G-AIRL algorithm can accurately simulate pedestrian behaviors during pedestrian-vehicle interactions and outperform all the baseline methods and ablation studies. Meanwhile, the trajectory and ARG weight visualization results demonstrate the importance of surrounding pedestrians under diverse interactions. Moreover, the analysis of the recovered reward function validates the “safety in numbers” effect and reveals the basic patterns of pedestrians interacting with vehicles without surrounding pedestrians and the heterogeneous impact of the size and morphologies of pedestrian groups on pedestrian behaviors. These results highlight the importance of incorporating group behaviors, offering insights into enhancing road safety and the efficiency of autonomous systems.