Paper figure sources

Figures are from the corresponding publications or the authors’ local manuscripts. Cropped PDF figures retain their content; thumbnails use CSS contain sizing. Manuscript versions are labeled on the paper page.

• Real-time lane-level abnormal traffic detection on freeways using sparse telematics data — Publication figure. Source. Fig. 1. Overview of the real-time abnormal traffic detection framework.

• Forecasting the Emergence and Evolution of Crash Hotspots: A Unified Deep Learning Framework for Proactive Traffic Safety — Publication figure. Source. Figure 1: Community Maps Predictive Analytics interface: (a) spatial hotspot heatmap; (b) filter controls; (c) hotspot detail popup.

• Geometry-Aware Infrastructure-Anchored Denoiser for UWB Sensing and Work-Zone Reconstruction — Publication figure. Source. Figure 1: Overview of the proposed geometry-first UWB denoising framework and motivating examples.

• CrashSight: A Phase-Aware, Infrastructure-Centric Video Benchmark for Traffic Crash Scene Understanding and Reasoning — Publication figure. Source. Figure 1. Overview of CrashSight-VQA. (a) Phase-aware temporal structure of a crash video. (b) VLM performance comparison across 7 QA categories (c) Example QA pairs spanning visual grounding and causal reasoning.

• Behavior-Grounded Lane Representation Learning for Multi-Task Traffic Digital Twins — Publication figure. Source. Figure 1 : Overview of the GeoLaneRep pipeline. Roadside observations are converted into per-lane geometry, trajectories, and descriptors (left). Three parallel encoders, spatial, temporal, and descriptor, produce per-lane representations that are fused via cross-lane multi-head attention into a shared semantic embedding space that supports multiple downstream tasks.

• An Agentic Workflow for Detecting Personally Identifiable Information in Crash Narratives — Publication figure. Source. Figure 1 : PII detection with the proposed agentic workflow.

• Real-time identification of cooperative perception necessity in road traffic scenarios — Publication figure. Source. Figure 1 from the paper.

• V2X-QA: A Comprehensive Reasoning Dataset and Benchmark for Multimodal Large Language Models in Autonomous Driving Across Ego, Infrastructure, and Cooperative Views — Publication figure. Source. Figure 1: Overview of V2X-QA. Left: representative examples of the twelve viewpoint-aligned tasks under vehicle-side, infrastructure-side, and cooperative settings. Right: MCQA-based training and evaluation pipeline of V2X-MoE on the V2X-QA training and testing splits.

• Geo-ORBIT: A Federated Digital Twin Framework for Scene-Adaptive Lane Geometry Detection — Publication figure. Source. Fig. 4: Overview of Knowledge-Based Lane Detection Algorithm. (a) Video detection and trajectory projection to GPS coordinates. (b) Lane center estimation using histogram analysis. (c) Lane-based trajectory clustering with KMeans. (d) Lane geometry estimation and boundary generation.

• SEAL: Vision-Language Model-Based Safe End-to-End Cooperative Autonomous Driving with Adaptive Long-Tail Modeling — Publication figure. Source. Figure 2: Overview of the SEAL framework. The vehicle-side and infrastructure-side views I v , I i I_{v},I_{i} are concatenated and processed by a frozen image encoder ℰ I \mathcal{E}{I} to produce visual token embeddings z z . The GMSAA module 𝒜 \mathcal{A} injects scenario-aware priors into z z based on scenario label d d . A trainable text encoder ℰ T \mathcal{E}{T} encodes the scene description E E into textual embeddings h h . The two modalities are fused via a transformer encoder ℰ \mathcal{E} , and decoded by 𝒟 \mathcal{D} to generate a future trajectory sequence Y ^ \hat{Y} .

• Goal-based neural physics vehicle trajectory prediction model — Publication figure. Source. Figure 1: Model architecture proposed in this paper. This dual sub-module framework first estimates the intentions and predicts multiple possible goals, then forecasts the full trajectories using a deep-learning enhanced social force model.

• AI-Enabled Digital Twin Framework for Safe and Sustainable Intelligent Transportation — Author manuscript figure. Source. Digital twin system overview.

• Sky-Drive: A distributed multiagent simulation platform for human–AI collaborative and socially aware future transportation — Author manuscript figure. Source. Fig. 1. Overview of Sky-Drive’s key components and functionalities. (a) a distributed multi-agent architecture enabling synchronized simulation across multiple terminals; (b) a multi-modal human-in-the-loop framework capturing comprehensive behavioral data through integrated sensor systems; (c) a digital twin framework that creates high-fidelity virtual replicas of transportation systems through multi-source data integration; (d) a human-AI collaboration mechanism facilitating knowledge exchange between humans and AI systems; (e) the planned integration of foundation models to enhance decision-making, enabling more adaptive and context-aware human-AI collaboration; (f) a hardware-in-the-loop framework, planned for future integration, ensuring that algorithms are evaluated in real-world environments.

• Planning Safety Trajectories with Dual-Phase, Physics-Informed, and Transportation Knowledge-Driven Large Language Models — Publication figure. Source. Fig. 1: LetsPi, a dual-phase, physics-informed LLM architecture for safe trajectory planning. The memory collection phase builds the knowledge database through in-depth reasoning and reflection using multiple safety metrics. The fast inference phase distills knowledge from the memory database, enabling timely trajectory planning.

• V2X-LLM: Enhancing V2X Integration and Understanding in Connected Vehicle Corridors — Publication figure. Source. Fig. 1: V2X-LLM Framework Architecture

• A Digital Twin Framework for Physical-Virtual Integration in V2X-Enabled Connected Vehicle Corridors — Publication figure. Source. Fig. 2. Three main components of the proposed digital twin system: the physical space (left), data pipeline (middle), and the digital space (right).

• A distributed deep reinforcement learning-based longitudinal control strategy for connected automated vehicles combining attention mechanism — Author manuscript figure. Source. Training process for the proposed control strategy.

• Enhancing Pedestrian Trajectory Prediction with Crowd Trip Information — Publication figure. Source. Figure 1: The concept of RNTransformer: pedestrian intention prediction with trip modality.

• Demystifying deep reinforcement learning-based autonomous vehicle decision-making — Publication figure. Source. Fig. 1: Architecture of the attention-based feature extractor.

• Real-World Data Inspired Interactive Connected Traffic Scenario Generation — Publication figure. Source. Figure 1: The Framework of The Interactive Traffic Scenario Generation Inspired by Real-World RSU Data

• Personalized lane departure warning based on non-stationary crossformer and kernel density estimation — Publication figure. Source. Figure 1 from the paper.

• Analyzing relationships between latent topics in autonomous vehicle crash narratives and crash severity using natural language processing techniques and explainable XGBoost — Publication figure. Source. Fig. 1. Research diagram of this study, the unstructured crash narratives are converted to collections of topics using topic modeling, and an XGBoost model is developed to estimate the relationships between topics and crash severity.

• How Does C-V2X Perform in Urban Environments? Results From Real-World Experiments on Urban Arterials — Publication figure. Source. Fig. 3. Locations of RSUs used for experiments and the setup of the OBU.

• Why did the AI make that decision? Towards an explainable artificial intelligence (XAI) for autonomous driving systems — Publication figure. Source. Figure 1 from the paper.

• A Probabilistic Framework for Estimating the Risk of Pedestrian-Vehicle Conflicts at Intersections — Publication figure. Source. Fig. 1. An example of pedestrian-vehicle conflicts.

• Real-Time Big Data Analytics and Proactive Traffic Safety Management Visualization System — Publication figure. Source. Fig. 1. System overview.

• A Hybrid Machine Learning Model for Predicting Real-time Secondary Crash Likelihood — Publication figure. Source. Fig. 2. Primary and secondary crash prediction model.

• Real-Time Crash Likelihood Prediction Using Temporal Attention–Based Deep Learning and Trajectory Fusion — Publication figure. Source. Fig. 1. Research flowchart.

• Improving Spatiotemporal Transferability of Real-Time Crash Likelihood Prediction Models Using Transfer-Learning Approaches — Publication figure. Source. Figure 1. Transfer-learning approaches used in this study. Note: LSTM = long short-term memory.

• Driving Maneuvers Detection using Semi-Supervised Long Short-Term Memory and Smartphone Sensors — Author manuscript figure. Source. Figure 1 from the paper.

• Using bus critical driving events as surrogate safety measures for pedestrian and bicycle crashes based on GPS trajectory data — Publication figure. Source. Fig. 1. Research approach flowchart.

• A Deep Learning Approach to Detect Real-Time Vehicle Maneuvers Based on Smartphone Sensors — Author manuscript figure. Source. Fig. 2. System workflow

• Prediction of pedestrian-vehicle conflicts at signalized intersections based on long short-term memory neural network — Publication figure. Source. Fig. 1. The studied locations.

• The application of novel connected vehicles emulated data on real-time crash potential prediction for arterials — Publication figure. Source. Fig. 1. Research area.

• Prediction of Pedestrian Crossing Intentions at Intersections Based on Long Short-Term Memory Recurrent Neural Network — Publication figure. Source. Figure 1. Spatial map of the studied site (30).

• Real-time crash risk prediction on arterials based on LSTM-CNN — Publication figure. Source. Fig. 2. Layout of Selected Segments.