본문으로 건너뛰기Skip to content
TUPA
Join TUPA

Research

교통 시스템과 도시를 데이터와 AI로 이해하고 설계합니다. 아래 다섯 축을 중심으로 연구합니다.We understand and design transport systems and cities with data and AI, organised around the five areas below.

01Areas

연구 분야Areas

Physical AI and Infrastructure-Centric AV Control

Physical AI·인프라 중심 자율주행 제어

자율주행차와 일반 차량이 섞인 혼재교통에서 차량 단독 지능을 넘어 인프라가 제어에 함께 참여하는 Physical AI 프레임워크를 연구합니다. 강화학습 기반 교차로·회전교차로 운영이 대표 주제입니다.We develop Physical AI frameworks in which infrastructure joins the control loop for mixed-autonomy traffic, including reinforcement-learning-based intersection and roundabout operations.

Physical AImixed-autonomy trafficreinforcement learningAV control

Crowd Dynamics

군중 동역학

보행자와 군중의 이동을 계측·모형화해 혼잡과 압사 위험을 예측하고 안전한 공간 설계를 지원합니다. VR 보행 시뮬레이터와 시뮬레이션 기반 군중 안전 분석을 수행합니다.We measure and model pedestrian and crowd movement to predict congestion and crush risk, supporting safer space design with VR walking simulators and simulation-based crowd safety analysis.

crowd dynamicspedestrian simulationcrowd safetyVR

LLM based Activity Based Model

LLM 기반 활동기반모형

대규모 언어모형(LLM)과 심층 생성모형으로 개인의 활동-통행 패턴을 합성하는 차세대 활동기반 수요모형을 연구합니다. 인구 합성과 자율주행 데이터 라벨링에도 LLM을 활용합니다.We build next-generation activity-based demand models that synthesize individual activity-travel patterns with LLMs and deep generative models, including population synthesis and AV data labeling.

LLMactivity-based modelpopulation synthesisgenerative models

Digital Twin Traffic Simulation

디지털 트윈 교통 시뮬레이션

미시교통 시뮬레이션과 디지털 트윈을 결합해 가상 환경에서 교통 운영과 자율주행을 검증합니다. KAIST 스핀아웃 dochak과 함께 드라이빙 시뮬레이터, 텔레드라이빙을 연구합니다.We integrate microscopic traffic simulation with digital twin technology to test operations and automated driving in virtual environments, together with the KAIST spinout dochak.

digital twinmicrosimulationdriving simulatorteledriving

Urban Science

도시 과학

20분 도시, 접근성·형평성, 공유 모빌리티-대중교통 연계 등 도시 스케일의 이동 현상을 데이터로 분석하고 설계합니다.We study city-scale mobility with data — the 20-minute city, accessibility and equity, and shared-mobility–transit integration.

urban scienceaccessibilityequity20-minute city

02Research Topics

연구 주제Research Topics

  • Traffic demand modeling using agent model

    Building a digital twin of road traffic with agent-based simulation (2022~2025): macroscopic transport simulation is integrated with activity-based modeling to reproduce realistic traffic flows and predict traffic volumes at the micro-analysis-zone level.

  • Real-time Prediction-based Conflict Measures for Evaluating Pedestrian Crossing Safety

    A unified prediction-driven pedestrian-vehicle safety evaluation approach for intelligent intersections, integrating deep learning-based behavior prediction with traffic conflict measures to evaluate pedestrian crossing risk in real time.

  • Human-centric traffic safety

    Investigating how drivers, pedestrians, and evacuees make decisions in safety-critical mobility situations through immersive VR experiments, covering micromobility safety, pedestrian-ADAS interaction, and underground/rail environments.

  • Analysis Public Transport Accessibility

    A comparative framework that measures the gap between planned transit schedules (GTFS) and actual smart-card usage with an extended G2SFCA accessibility model, diagnosing where public transport supply and real travel demand diverge.

  • Goal-Driven Experimental Pipeline for Multi-Vehicle Autonomous Driving in CARLA in ROS2

    A CARLA-ROS2 experimental pipeline for goal-driven autonomous driving on Lanelet2 maps, composing multi-vehicle scenarios that mix custom driving policies with rule-based autopilot vehicles to study interaction-aware driving behaviors.