Add BrightSurf on Google Email

UMass Amherst researchers identify top risk factors for pedestrian-vehicle crashes at Massachusetts bus stops

Researchers used machine learning to evaluate 1,773 bus stops across Massachusetts, identifying 13 crash-relevant characteristics and proposing countermeasures. The study found that bus stop types in mixed-use arterial corridors and dense urban cores have the highest prevalence of crashes.

SourceUniversity of Massachusetts Amherst·JournalData Science for Transportation·TypeData/statistical analysis·DateJan 22, 2026

From global open multi-source data to network-wide traffic flow: A large-scale case study across multiple cities

A team of researchers proposes a novel attention-based graph neural network to estimate network-wide traffic flow across multiple cities. The model leverages correlations between road traffic flow and urban characteristics, such as building structures, human activity, infrastructure connectivity, and dynamic traffic conditions.

SourceTsinghua University Press·JournalCommunications in Transportation Research·DateDec 23, 2025

How did COVID-19 change urban traffic?

Researchers at National University of Singapore used machine learning models to predict traffic congestion in Alameda County before, during, and after the pandemic. The study found that Bi-LSTM models were more accurate than traditional forecasting tools and provided transparent insights into the drivers of congestion changes.

SourceTsinghua University Press·JournalCommunications in Transportation Research·DateDec 23, 2025

How to make AI truly scalable and reliable for real-time traffic assignment?

A new framework, MARL-OD-DA, offers a promising answer to making AI truly scalable and reliable for real-time traffic assignment. The approach redesigns learning agents at the origin–destination level, utilizing Dirichlet-based continuous actions to achieve stable and high-quality solutions under dynamic travel demand.

SourceTsinghua University Press·JournalCommunications in Transportation Research·DateDec 23, 2025

Seeing farther: A new camera-based technique detects distant vehicles for safer roads

A new method analyzes nearby vehicle motion to estimate road's trajectory and vanishing point, capturing distant road areas and enhancing safety. The system outperformed conventional and deep learning-based techniques in tests under day and night conditions, reducing intersection-related accidents.

SourceShibaura Institute of Technology·JournalIEEE Open Journal of Intelligent Transportation Systems·TypeComputational simulation/modeling·DateDec 15, 2025

Towards integrated data model for next-generation bridge maintenance

Researchers develop a novel integrated data model that merges construction and geospatial information standards to manage bridges' 3D geometry data and maintenance records. This framework enables accurate damage location assessment, repair prioritization, and predictive maintenance, leading to improved infrastructure safety and longevity.

SourceHosei University·JournalComputer-Aided Civil and Infrastructure Engineering·TypeComputational simulation/modeling·DateNov 27, 2025

Cheaper cars pollute more than expensive cars, leading to emissions inequality

Research by University of Birmingham scientists reveals that lower-income individuals are more likely to own cheaper, higher-emitting vehicles contributing disproportionately to local urban air pollution. Spending an additional £10,000 on a diesel vehicle is associated with a 40% reduction in nitrogen oxide emissions per litre.

SourceUniversity of Birmingham·JournalJournal of Cleaner Production·TypeData/statistical analysis·DateNov 14, 2025

The next frontier in clean flight? Jet fuel from city waste

Researchers explore using municipal solid waste as a low-emission, cost-effective feedstock for sustainable aviation fuel, reducing greenhouse gas emissions by 80-90%. The study suggests that adopting municipal solid waste-based jet fuels could save airlines money under carbon pricing systems.

SourceHarvard John A. Paulson School of Engineering and Applied Sciences·JournalNature Sustainability·TypeData/statistical analysis·DateNov 12, 2025

PSU research shows Portland transit-oriented developments reduce car trips, especially at affordable housing sites

A new study by Portland State University's Transportation Research and Education Center (TREC) reveals that transit-oriented developments (TODs) in the Portland metro area generate significantly fewer car trips than previously estimated. At affordable housing sites, car trips were reduced to only a quarter to two-fifths of expected rates.

Worcester Polytechnic Institute selected as key partner in national cybersecurity and AI training initiative to advance U.S. automotive innovation

The $2.5 million DRIFT program aims to close critical talent gaps and safeguard connected vehicles by providing specialized online and in-person training. Worcester Polytechnic Institute will lead the university’s DRIFT program, offering tuition-free modules and real-world training to upskill engineers and professionals.

DFUN-KDF: A knowledge distillation-based decentralized federated learning framework for uav network optimization

The DFUN-KDF framework uses federated knowledge distillation to enable UAVs to extract information from embedded data and update their local models, reducing energy consumption and improving adaptability. The framework's filtering mechanism eliminates biased embeddings, ensuring the stability and reliability of the system.

SourceTsinghua University Press·JournalCommunications in Transportation Research·DateJul 21, 2025

New study: Teen drivers safer with more practice

A new study by Virginia Tech Transportation Institute found that teen drivers who practiced more had 30% fewer crash or near-crash incidents. Researchers analyzed data from 82 teen drivers monitored for 22 months, revealing that supervised driving experiences and safety education are crucial for reducing novice driver risks.

Bridging microscopic interactions and macroscopic traffic patterns: a novel approach to stochastic fundamental diagram modeling

Researchers propose a novel framework to model stochastic fundamental diagrams from microscopic interactions, deepening understanding of traffic flow's stochasticity. The Leader-Follower Conditional Distribution-based Stochastic Traffic Modeling (LFCD-STM) framework offers high consistency with real-world data and has implications for ...

SourceTsinghua University Press·JournalCommunications in Transportation Research·DateMar 13, 2025

Researchers from Incheon National University advanced adaptive traffic monitoring with smart cameras

The researchers developed a novel camera-based system that adapts to traffic flow in real-time, ensuring efficient monitoring and resource use. The system's two approaches, the Random-Value-Camera-Level Algorithm and the ALL-Random-With-Weight Algorithm, optimize camera usage and save energy while maintaining reliable surveillance.

SourceIncheon National University·JournalIEEE Internet of Things Journal·TypeExperimental study·DateJan 15, 2025