The transition to Electric Vehicles (EVs) demands intelligent, congestion-aware infrastructure planning to balance user convenience, economic viability, and traffic efficiency. We present a joint optimisation framework for EV Charging Station (CS) placement and pricing, explicitly capturing strategic driver behaviour through coupled non-atomic congestion games over road networks and charging facilities. From a Public Authority (PA) perspective, the model minimises social cost, travel times, queuing delays and charging expenses, while ensuring infrastructure profitability. To solve the resulting Mixed-Integer Nonlinear Programme, we propose a scalable two-level approximation method, Joint Placement and Pricing Optimisation under Driver Equilibrium (JPPO-DE), combining driver behaviour decomposition with integer relaxation. Experiments on the benchmark Sioux Falls Transportation Network (TN) demonstrate that our method consistently outperforms single-parameter baselines, effectively adapting to varying budgets, EV penetration levels, and station capacities. It achieves performance improvements of at least 16 % over state-of-theart approaches. A generalisation procedure further extends scalability to larger networks. By accurately modelling traffic equilibria and enabling adaptive, efficient infrastructure design, our framework advances key intelligent transportation system goals for sustainable urban mobility.
Interpreting ongoing but incomplete current driving actions and anticipating future driving intentions are formulated as Online Action Detection (OAD) and Online Action Anticipation (OAA). However, existing methods generally follow unidirectional forward temporal modeling, which imposes a forward constraint that prevents subsequent frames from correcting earlier modeling errors, leading to the accumulation of misperceptions in driving behavior. To this end, we propose to incorporate backward temporal modeling as a complementary strategy following forward temporal modeling. The backward temporal modeling mitigates the effect of early-stage misinterpretation through reevaluating and reinterpreting earlier ambiguous cues based on subsequent context. Based on this, we propose a unified model for OAD and OAA, named Bidirectional Temporal Perceiver (BiTPerceiver). Specifically, BiTPerceiver extracts task-relevant information as video memory from online videos via Transformer Decoders. Then, motivated by the recent success of Mamba in sequence modeling, BiTPerceiver models forward-then-backward temporal dependency in the video memory through Mamba. The final modeling result includes current and future action representations. BiTPerceiver achieves state-of-the-art OAD and OAA performance on the Honda Research Institute Driving Dataset (HDD) (OAD: 41.3%, OAA: 27.3%), THUMOS'14 (OAD: 72.6%, OAA: 59.3%) and TVSeries (OAD: 89.8%, OAA: 83.6%).
This paper presents an optimal pre-signal-based traffic signal framework that enhances intersection traffic flows while enabling efficient driving behavior, suitable for traditional and automated vehicles (AVs). Specifically, we propose the inclusion of an auxiliary traffic signal at an early stop line located before the standard one at the intersection, which allows the leading vehicle to enter the intersection immediately at a green light with a higher speed, reducing the start-up loss time by activating the auxiliary green light slightly earlier than the main conventional traffic signal. The relative positioning of the early stop line is optimized numerically for effectively redesigning the signaling system, considering relevant safety issues, constraints, and objectives. The vehicle control system employs model predictive control (MPC) to minimize costs associated with velocity and acceleration while maintaining a safe following distance. Through microscopic traffic simulations, the performance of traditional and automated vehicles is analyzed within this new signal system. The findings indicate a marked improvement in average speed, travel duration, fuel efficiency, and idling time compared to conventional intersection traffic management.
Maintaining high-quality road infrastructure is essential for ensuring traffic safety and minimizing accidents. Consequently, developing a low-cost and efficient road pothole detection and repair system is important to improve road quality and travel safety. In recent studies, vibration signals from cell phone sensors are used for detection. The existing methods usually suffer from high computational complexity and lack of contextual information extraction. To solve these problems, this study proposes a network architecture based on the Inceptiontime framework, PCAG-Inception. Firstly, instead of the conventional operation, we use the ICAM (Inception-based Cross-Attention Module) to operate on only some channels to improve the inference speed. Then, we introduce a cross-attention mechanism with shared weights to establish the relationship between features extracted by convolution at different scales, thus learning a more fine-grained representation. In addition, to better aggregate temporal contextual information, we propose a new gated temporal attention unit. This module captures both local and global features in a time series through a gating mechanism. The gate dynamically adjusts feature weights, enhancing the model's ability to capture interactions between different variables. In this study, a series of experiments are conducted on our constructed vibration signal dataset of road potholes, and the results show that our proposed PCAG-Inception model achieves excellent performance with an accuracy of 94.51 %.
Accurate traffic flow prediction is a critical component of intelligent transportation systems. While previous studies have made significant breakthroughs in modeling spatio-temporal correlations, existing methods still suffer from two significant limitations: (i) underdetermined traffic flows caused by sparse sensor deployment, where incomplete path observations make it difficult to capture non-local spatial dependencies. (ii) nonequilibrium traffic flows arising from congestion propagation and varying driver reaction times, which induce asynchronous temporal correlations and nonlinear dependencies. To address these challenges, we propose a novel Generative Adversarial Network and Large Language Model-based Adaptive SpatioTemporal framework (GLAST). Specifically, It employs an adaptive edge weight augmentation mechanism based on GAN with residuals to capture deviations in periodic patterns and construct dynamic adjacency matrices, which enables the capture of nonlocal spatial dependencies under sparse sensing conditions. The weighted attention mechanism is applied to aggregate lagged adjacency matrices to model asynchronous temporal correlations, effectively capturing temporal delays and nonlinear dependencies. Furthermore, adaptive temporal descriptors and congestion metrics are encoded as prompts for the LLM to enhance the modeling of multi-scale spatio-temporal relationships. Experimental results on benchmark datasets demonstrate that GLAST consistently outperforms various state-of-the-art baselines. Given the widespread presence of sparse sensing and congestion propagation in real-world highways, the proposed framework also holds the potential to inspire other urban traffic applications.
Adaptive traffic signal control (TSC) has demonstrated strong effectiveness in managing dynamic traffic flows. However, conventional methods often struggle when unforeseen traffic incidents occur (e.g., accidents and road maintenance), which typically require labor-intensive and inefficient manual interventions by traffic police officers. Large Language Models (LLMs) appear to be a promising solution thanks to their remarkable reasoning and generalization capabilities. Nevertheless, existing works often propose to replace existing TSC systems with LLM-based systems, which can be (i) unreliable due to the inherent hallucinations of LLMs and (ii) costly due to the need for system replacement. To address the issues of existing works, we propose a hierarchical framework that augments existing TSC systems with LLMs. This framework leverages LLMs at the upper level to dynamically fine-tune selected parameters of signal controllers at the lower level in response to real-time traffic incidents. To enhance domain-specific reliability, we employ Retrieval-Augmented Generation (RAG) to draw knowledge from a tailored traffic language database that encompasses traffic conditions and controller operation principles. Our results show that LLMs can serve as trustworthy virtual traffic police officers that can adapt conventional TSC methods to traffic incidents with significantly improved operational efficiency and reliability.
Efficient real-time traffic management during evacuations is critical for mitigating the impacts of disasters. This study proposes real-time contraflow operations formulated as a discrete-variable Model Predictive Control (MPC) problem. To address the combinatorial complexity, we develop a novel sampling-based optimization algorithm that integrates Zero-suppressed binary Decision Diagrams (ZDD) with the crossentropy method. The macroscopic fundamental diagram is employed to aggregate zone-based traffic dynamics, enabling rapid state prediction. Numerical experiments show that the ZDD-based sampling achieves 2.2 to 4.7 times faster computation compared to a baseline method. A case study in Tokyo demonstrates that the proposed controller keeps traffic below critical congestion thresholds and improves trip completion by approximately 15 % relative to the scenario without contraflow. These results underscore the potential of discrete-variable MPC with ZDD acceleration as a practical decision-support tool for dynamic traffic management during disasters.
Simulation is considered a critical component in the validation of autonomous driving systems, particularly for evaluating rare and safety-critical scenarios. While recent trajectory prediction methods have shown strong capabilities in generating realistic future trajectories, these models are typically designed to estimate the most likely outcome and do not support explicit conditioning on scenario parameters. As a result, their use in targeted simulation or scenario design remains limited. To address this, a framework for conditioned trajectory generation is presented, enabling the generation of a set of diverse and physically plausible trajectories from a single initial scenario state, conditioned on interpretable highlevel variables such as desired endpoints or average velocity. A hybrid machine learning framework is employed to generate control actions, which are subsequently processed by a kinematic motion model. Expert knowledge is incorporated through soft physical constraints and a smoothness loss, ensuring physical realism and signal continuity. Through experiments, it is demonstrated that realistic and diverse trajectories aligned with scenario context can be produced. The proposed framework can be integrated into simulation environments for validating autonomous vehicles.
Efficient taxi repositioning is essential to address the spatiotemporal mismatch between urban travel demand and vehicle supply. Traditional approaches often rely on heuristic rules or reinforcement learning, which suffer from limited generalization and poor interpretability. In this paper, we propose an LLM-driven framework for intelligent taxi repositioning that integrates three key components: (1) LLM-based decision reasoning, (2) spatiotemporal clustering for demand pattern abstraction, and (3) experience-driven responsibility assignment for adaptive policy optimization. A city-scale simulator is developed to evaluate the effectiveness of our approach. Experimental results show that our method achieves the highest response rate of 70.59%, the lowest average pickup time of 153.3 seconds, and the highest vehicle utilization rate of 48.94%, outperforming four baselines including clustering and optimization-based methods. Our simulator and code are publicly available at:https://github.com/cloverSaber/TrafficSimulator.
Traffic Accident Detection (TAD) in surveillance videos is a critical task in Intelligent Transportation Systems (ITS). However, current TAD does not analyze the fine-grained information of the specific accident, only identifies the existence or occurrence time of traffic accidents in a video. This study presents a novel Dataset named STTAD that covers fine-grained information such as multiple categories and their Spatial Temporal Occurrence Regions in surveillance videos. Moreover, a tailored deep learning algorithm named STFN is proposed for the implementation of Event-Level TAD. Experimental results demonstrate that STFN could effectively extract the video features and detect the Spatial Temporal Occurrence Regions of multiple accident categories, but further efforts are indeed needed in Event-Level TAD. The STTAD dataset and the tailored algorithm will be open-sourced for research use available through https://github.com/ZTR02/STTAD.git.
Motorized two-wheelers (MTW) dominate Indian roads but remain underrepresented in driver behavior research. This study presents the first large-scale analysis of MTW driver gaze behavior in naturalistic, heterogeneous urban traffic, using the myEye2Wheeler dataset. A semantic segmentation pipeline (YOLOv11 + SAM2) was used to extract object-level gaze metrics under two attention modes: direct gaze (foveal overlap) and central vision (parafoveal monitoring). Results reveal a functional division: central vision supports broad monitoring, while direct gaze enables brief, selective sampling. Novice riders exhibit road-anchored scanning, returning to the road between object fixations, while experienced riders form longer chains of attention across multiple objects. The findings suggest that experience primarily refines temporal rhythm rather than altering allocation strategy and reduces object-class effects in gaze patterns. These findings offer new insight into MTW attention structures and inform future work on behavior modeling and safety systems.
This study provides a semi-analytical solution to the charging optimization problem for electric vehicles, using Pontryagin's Maximum Principle (PMP) to derive explicit expressions for optimal control and costate trajectories. Such expressions enable much faster trajectory computation than numerical optimization, making this method suitable for larger trip-planning problems. In addition, these explicit expressions reveal the fundamental structure of the optimal solution and how all the parameters and variables affect it, leading to more interpretable results and deeper theoretical insights than fully numerical or black-box methods allow. The method nearly eliminates online computation, while having an average difference with a benchmark solution of about 2.74%.
This paper presents an integrated framework for autonomous mobility-on-demand (AMoD) systems, focusing on dynamic pricing, ride-sharing, and decentralized coordination. Built on a high-fidelity, city-scale environment calibrated with NYC taxi data, the framework dynamically generates passenger and shared autonomous vehicle (SAV) agents based on realworld spatiotemporal demand patterns. The system integrates a multi-objective multi-agent deep reinforcement learning (MOMADRL) framework with centralized training and decentralized execution (CTDE), allowing agents to optimize individual incentives and system-level social welfare jointly. Adaptive pricing strategies, flexible ride-matching mechanisms, and zone-based geographic abstractions are included to enhance computational efficiency while maintaining geographic realism. Experimental results demonstrate that our framework consistently improves key performance indicators like passenger waiting times, vehicle utilization, and pricing stability, outperforming purely centralized or decentralized methods. This research provides a robust platform for testing adaptive policies and advancing scalable, equitable AMoD system design.
Precise understanding of lane-level road topology and the semantic relationships between traffic elements and lanes is crucial for highly autonomous driving systems, especially for ensuring safe and compliant navigation through complex road networks. Current methods for lane graph generation primarily focus on geometric and structural aspects, largely neglecting the critical semantic associations to traffic lights and signs. To address this gap, we propose MRLGraph, a novel graph-based neural network method that explicitly learns and predicts the missing map relations using geometric lane-level map information. Our approach utilizes graph attention mechanisms and path-based encoding of the lane segments to accurately capture both local geometric context and broader topological relationships. In extensive experiments, we demonstrate that our method significantly outperforms existing approaches and baseline models in terms of precision, recall, and generalization capabilities both on the existing Lyft and our own MRLScenes dataset. The latter is a novel hybrid dataset combining real-world high-definition map samples and diverse synthetic scenarios. We will make the code of MRLGraph and the MRLScenes dataset available with the publication of this paper.
Traffic monitoring is crucial for effective traffic management, planning, and safety assessment. Conventional monitoring methods depend heavily on fixed infrastructure such as loop detectors and stationary video cameras, which, despite their accuracy, have inherent limitations including restricted spatial coverage, susceptibility to adverse weather conditions. In contrast, the rise of autonomous vehicles (AVs) equipped with advanced sensors such as LiDAR presents a unique opportunity. Although primarily intended for vehicle navigation and safety, these sensors continuously collect rich datasets that inherently provide comprehensive traffic information. This study proposes a proof-of-concept framework to demonstrate the feasibility of extracting detailed traffic metrics—such as lane-level speed, density, traffic flow, and time-to-collision (TTC)—from vehicle-mounted LiDAR sensors. Using the publicly available Waymo Open Dataset and the MCTrack algorithm, the research highlights how individual AV sensor data can be leveraged to derive meaningful traffic indicators without requiring additional public sector infrastructure investment. Furthermore, this study considers a scenario in which aggregating multiple AVs' sensor data significantly enhances spatial coverage and improves the reliability and comprehensiveness of macro-level traffic metrics as AV market penetration increases.
Center-based 3D object detectors have become increasingly popular due to their advantageous balance between speed and accuracy. However, by relying on standard 2D Gaussian distributions, these methods overlook the representation of object shape and rotation in their learning objectives. This presents challenges to the stability of object detection in practical applications. To address the issue, this work proposes a more generalized Gaussian representation of objects, upon which a novel center-based 3D object detection framework named CenterNext is introduced. Specifically, the proposed method designs a generalized 2D Gaussian kernel to construct object heatmaps. By preserving the unambiguous characteristics of the invariant object center and incorporating variations in object shape and rotation into the formulation, the biases introduced by the previous insufficient object representation are mitigated. The proposed method effectively guides models to recognize the distinctiveness of objects during optimization, thereby enhancing the generalization performance of 3D object detection. Extensive experiments conducted on KITTI, Waymo, and ONCE demonstrate the effectiveness of our approach. With the newly proposed CenterNext, the existing center-based detectors are elevated to a higher level.
Distributed drive electric vehicles (DDEV) employ multiple in-wheel or wheel-side motors as power sources. Due to the direct connection between the motors and the wheels, and the elimination of traditional clutches, the introduction of motors alters the vehicle's inherent damping characteristics through their interaction with components such as the suspension and shock absorbers. Specifically, during Anti-lock Braking System (ABS) intervention, the wheels are susceptible to high-frequency oscillations, which may trigger frequent torque reductions by the ABS. This, in turn, degrades braking performance and may even lead to severe issues, such as the loss of braking force. This paper proposes an anti-oscillation control strategy based on Electro-Mechanical Braking (EMB). First, a tire model is developed that incorporates the damping effects of the motors and the suspension system, and the system's resonance peak frequency is analyzed. Second, a high-frequency oscillation detection method is designed based on wheel speed characteristics. Finally, a Model-Based Predictive and Logic Control (MPLC) framework is introduced. This framework suppresses highfrequency oscillations by regulating the amplitude and frequency of ABS torque reductions, thereby ensuring sufficient braking force. Simulation results demonstrate that the proposed method significantly reduces braking distance and improves braking deceleration. Compared to conventional ABS algorithms, the proposed strategy achieves substantial improvements in braking performance.
Autonomous Intersection Management (AIM) systems coordinate Connected and Autonomous Vehicles (CAVs) without traffic signals. These systems offer the potential for improved safety and efficiency. However, their scalability remains a major challenge, especially in determining the appropriate spatial extent of coordination, which is referred to as the control range and strongly affects computational cost. This paper investigates how both the design of the control range and the strategy for selecting vehicles to re-plan influence the trade-off between coordination performance and computation. We define the control range as a tunable parameter and propose a selection method that focuses only on vehicles that are likely to interact through potential collisions. Through simulation under various traffic densities, ranging from 5 to 20 vehicles, and different control range lengths from 25 to 150 m, we show that the proposed method approximately reduces intersection delays by up to 30 % and lowers computation time by as much as 90 %. Furthermore, we observe that both metrics tend to saturate when the control range exceeds approximately 100 m. These findings suggest that scalable coordination can be achieved by adaptively choosing which vehicles to re-plan and how far ahead to consider, based on traffic conditions.
Imbalanced datasets pose a significant challenge in machine learning, where rare situations can significantly impact model performance, particularly in safety-critical applications such as autonomous driving. To ensure models generalize well to new data and real-world scenarios, it is essential to address imbalances during the dataset creation. We propose a novel approach to improve the distribution of human pose datasets by detecting and mitigating rare poses in human motion data. Our method, Out-Of-Distribution Mitigation based on Dynamic Sampling (OODM-DS), employs dimensionality reduction to transform high-dimensional human pose data into a low-dimensional latent space, facilitating efficient analysis of human poses. By analyzing the latent space using density estimation, we propose a dynamic sampling strategy to generate a dataset with a connected simulation. This approach significantly improves the performance of a human pose estimation model on rare poses, while also reducing PCKh metric variance between rare and non-rare poses, making it a promising tool for dataset production in applications such as autonomous driving where safety-critical data can be rare and difficult to obtain.
