Understanding the impacts of negative advanced driving assistance system warnings on hazardous materials truck drivers’ responses using interpretable machine learning

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 146, P. 110308 - 110308

Published: Feb. 20, 2025

Language: Английский

---

A Virtual Vehicle–Based Car‐Following Model to Reproduce Hazmat Truck Drivers’ Differential Behaviors

Journal of Advanced Transportation, Journal Year: 2024, Volume and Issue: 2024(1)

Published: Jan. 1, 2024

Enhancing hazmat truck safety through advanced driving assistance systems (ADAS) relies on both system efficacy and driver reactions. This study investigates the behaviors of drivers in response to forward collision warnings (FCWs). Traditional warning triggering methods struggle capture diverse immediate responses; therefore, our research employs a vision‐based framework for data extraction utilizes K‐means++ clustering method response‐based classification. Moreover, we propose an enhanced version intelligent model (IDM) based concept virtual vehicle reproduce drivers’ differential during risky car‐following periods, achieving results that depict improved simulations. is compared with classic benchmarks, including IDM, optimal velocity (OVM), full difference (FVD) model, demonstrating superior performance terms traffic stability extreme scenarios. Our findings highlight preaction tend accelerate before receiving warnings, opting overtake rather than maintain safe distances. In contrast, calm decelerate anticipation warning, showcasing their awareness maintaining safety. The analysis reveals aggressive are predominantly 41–45 age group, indicating higher skill level, while more commonly older, reflecting trend cautious behaviors. Overall, contributes development effective ADAS by considering real‐time responses emphasizes potential revolutionize commercial adoption enhance road operations.

Language: Английский

---

Dynamic Evaluation Method for Public Transport Operation Indicators Based on Rank Transformation

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: Feb. 21, 2025

<div class="section abstract"><div class="htmlview paragraph">This study presents a method to evaluate the daily operation of traditional public transportation using multi-source data and rank transformation. In contrast with previous studies, we focuses on dynamic indicators generated during vehicle operation, while ignoring static indicators. This provides better reference value for management transport vehicles. Initially, match on-board GPS network stop coordinates extract arrival departure timetable. helps us calculate operational metrics such as dwell time, interval, frequency bunching large interval. By integrating IC card timetable, can also estimate number people boarding at each derive passenger waiting average time. Finally, developed comprehensive evaluation performance, covering three dimensions: bus stops, vehicles, routes. uses K-means clustering classify applies transformation techniques score. At levels, use principal component analysis(PCA) identify key influencing factors, anf apply service-level classification. route level, perform time frequency. Delphi is used determine relative weights indicator, so facilitate ranking routes according applicable 20 in Shenzhen, involving 293 vehicles 506 stops. The results show that this effectively make contribution management.</div></div>

Language: Английский

---

Construction and Application of Traffic Accident Knowledge Graph Based on LLM

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: Feb. 21, 2025

<div class="section abstract"><div class="htmlview paragraph">Records of traffic accidents contain a wealth information regarding accident causes and consequences. It provides valuable data foundation for analysis. The diversity complexity textual pose significant challenges in knowledge extracting. Previous research primarily relies on Natural Language Processing (NLP) to extract from texts uses graphs (KGs) store structured way. However, the process based NLP typically necessitates extensive annotated datasets model training, which is complex time-consuming. Moreover, application by direct querying within graph requiring commands, leads poor interaction capabilities. In this study, we adapt an innovative approach integrates Large Models (LLMs) construction graph. Based defined schema layer graph, employ LLMs records refine extraction using prompts few-shot learning mechanism. To ensure accuracy extracted result, dual verification method combines self-verification with manual inspection. Then visualize Neo4j. Finally, explore KGs framework Retrieval-Augmented Generation (RAG) construct intelligent question-answering system. combination facilitates semi-automated Knowledge Graph-Based Question Answering System Traffic Accidents enables query answering tasks such as causation analysis scenario generation autonomous driving tests. integration not only expands scenarios but also reduces risk hallucination responses generated LLMs. This efficiently Extracting unstructured data, advances digitalization intelligence management.</div></div>

Language: Английский

---

Improvements aiming at a safer living environment by analyzing crash severity through the use of boosting-based ensemble learning techniques

International Journal of Crashworthiness, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 16

Published: April 22, 2025

Language: Английский

---

Enhancing road safety with machine learning: Current advances and future directions in accident prediction using non-visual data

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 137, P. 109086 - 109086

Published: Aug. 11, 2024

Language: Английский

---

Freeway Traffic Conflict Forecasting: A Machine Learning Approach with RF-LSTM Integration

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: March 19, 2025

<div class="section abstract"><div class="htmlview paragraph">This paper aims to forecast and examine traffic conflicts by integrating Random Forest (RF) alongside Long Short-Term Memory Network (LSTM). The begins with the method, pinpointing essential elements affecting conflicts, revealing that speed difference between interacting vehicles their leaders, as well average headway distance have significant effects on occurrence of conflicts. forecasted Time Collision (TTC) metric demonstrates extraordinary accuracy, confirming creation a precise conflict model. model expertly predicts vehicle's trajectory. This skillfully anticipates vehicle paths potential conflict, demonstrating strong alignment actual patterns offering support for management highlighting imminent risks. Merging RF feature selection LSTM temporal dynamics enhances forecasting capability. Furthermore, it also illuminates changes in interaction patterns. Considering both fixed shifting elements, this extensive process leads deep understanding subtle mechanisms driving suggested platform serves robust device engineers policymakers, enabling them make informed decisions implement effective strategies managing traffic.</div></div>

Language: Английский

---

A Real-Time Multiplayer SUMO-Unity Co-Simulation Platform for Vehicle-Aircraft Coordinated Operation

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: Feb. 21, 2025

<div class="section abstract"><div class="htmlview paragraph">As the demands for air travel and cargo continue to grow, airport surface operations are becoming increasingly congested, elevating operational risks all entities. Conventional measurement methods in traffic scenarios limited by high temporal spatial costs, uncontrollable variables, their inabilities account low-probability events. Moreover, current simulation software exhibits weak capabilities poor interactivity. To address these issues, this study developed a virtual reality platform operations. The integrated 3D modeling technologies, including Blender Unity, with Photon Fusion multiplayer Simulation of Urban Mobility (SUMO) software. By incorporating Logitech external devices, enabled real-time human-driven simulations, online interactions, validation flow models. enhance practical applicability platform, scenario library vehicle-aircraft-taxiway coordinated was designed based on historical data. A stated preference survey distributed aviation experts, evaluating risk ratings occurrence frequencies. Principal component analysis rank sum ratio were applied identify key scenarios, which embedded into platform. results simulate interaction among vehicles, aircraft, taxiways, providing scenario-driven control strategy verification interactive driving decision support. This approach contributes digital transformation management, enhancing efficiency safety.</div></div>

Language: Английский

---

Analysis of Crash Precursors under Different Traffic Flow Conditions with Enhanced Interpretability

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: Feb. 21, 2025

<div class="section abstract"><div class="htmlview paragraph">This study investigates the precursors of crashes under varying traffic states through an in-depth analysis freeway data. This method effectively addresses limitations associated with using surrogate measures in safety research. We used k-means clustering to categorize into three types: free flow, transitional state, and congested flow. By employing case-control experimental approach, we conducted During feature selection process, set matching rules choose control group data that meet criteria time, location, state. Initially, flow variables were constructed based on multiple dimensions, including time window width, spatial parameters, statistical characteristics. To reduce multicollinearity, correlation matrices variance inflation factors (VIF). then applied Recursive Feature Elimination (RFE) combined XGBoost model select key features, interpreted impact these features crash occurrence SHapley Additive exPlanations (SHAP) value. Finally, employed a logistic regression evaluate selected important reflecting relationship between from broad perspective. The results indicate significant differences main affecting different conditions. In variability speed is more significant. vehicle distribution across lanes significantly affect crashes; while standard deviation speeds among upstream average downstream have greater crashes. not only enhances interpretability methods but also provides basis for management departments formulate corresponding strategies scenarios.</div></div>

Language: Английский

---

Lightweight Algorithm Optimization of Visual SLAM on Embedded Computing Devices for Intelligent Transportation Systems

SAE technical papers on CD-ROM/SAE technical paper series, Journal Year: 2025, Volume and Issue: 1

Published: Feb. 21, 2025

<div class="section abstract"><div class="htmlview paragraph">In the context of intelligent transportation vehicle perception, embedded computing devices serve as primary platform, facing challenge traditional visual SLAM(Simultaneous Localization and Mapping) framework's high computational demands for environmental feature points. To address issues such point cloud drift errors in long-term, large-scale road traffic perception tasks mismatch rate tracking scenes with numerous dynamic objects, this work proposes an optimized elimination method odometry module based on ORB-SLAM3 framework. Additionally, efficient vector dictionary loading matching algorithm repetitive keyframes is designed loop closure detection module. In calculation module, a confidence index introduced to eliminate mismatched points objects. Meanwhile, binary applied optimize vocabulary matching, addressing scene re-localization problem during environment detection. The was tested evaluated KITTI dataset using RDK X3 device. Results indicate that framework maintains accuracy map coordinate calculations without degradation achieves real-time performance reconstruction scenes. Moreover, speed memory usage are superior original SLAM framework.</div></div>

Language: Английский

---