Researchers at Tsinghua University developed PriorFusion, a unified framework that integrates semantic, geometric, and generative shape priors to significantly improve the accuracy and stability of road element perception in autonomous driving systems. The research addresses a long-standing challenge: existing end-to-end perception models often generate irregular shapes, fragmented boundaries, and incomplete road elements in complex urban scenarios.

Ride-pooling is widely recognized as a sustainable way to ease congestion, reduce costs and cut emissions, yet adoption remains limited.  When operators act independently, efficiency is low because requests cannot be matched across platforms.  Aggregation platforms seek to improve this by forcing all operators into a permanent coalition, but differences in size, cost and market position make such arrangements unstable.  To address this, researchers from Beihang University and Delft University of Technology developed a multi-level coalition formation game framework that enables coalitions to form dynamically in response to trip requests, allowing flexible cooperation without requiring all operators to remain in a single group at all times.

To answer this question: How to make AI truly scalable and reliable for real-time traffic assignment? A research team from KTH Royal Institute of Technology, Monash University, Technical University of Munich, Southeast University, and the University of Electro-Communications has developed a new framework—MARL-OD-DA—that offers a promising answer. The approach redesigns learning agents at the origin–destination (OD) level and utilizes Dirichlet-based continuous actions to achieve stable and high-quality solutions under dynamic travel demand.

A “standard reference thermoelectric module (SRTEM)*” for objectively measuring thermoelectric module performance has been developed in Korea for the first time. A research team led by Dr. Sang Hyun Park at the Korea Institute of Energy Research (KIER; President Yi, Chang-Keun) developed the world’s second standard reference thermoelectric module, following Japan, and improved its performance by more than 20% compared with existing modules, demonstrating the excellence of Korea’s homegrown technology.

To address the trade-off between accuracy and cross-city generalization in traffic flow estimation, a research team from The Hong Kong Polytechnic University and New York University proposes a novel framework based on global open multi-source (GOMS) data, including urban structures and population density. By developing an advanced graph neural network model that effectively fuses these static urban features with dynamic traffic data, the study achieves stable and accurate network-wide traffic estimation, as validated across 15 diverse cities in Europe and North America.

Researchers at National University of Singapore used multiple interpretable machine learning methods to predict traffic congestion in in Alameda County in the San Francisco Bay Area, USA, during the pre-lockdown, lockdown, and post-lockdown periods.