Intelligence & Interaction Lab - Kookmin
Transcript of Intelligence & Interaction Lab - Kookmin
Intelligence and Interaction Lab Kookmin University
Intelligence & Interaction Lab 지능 인터랙션 실험실
Graduate School of Automotive EngineeringKookmin University
Seoul, Korea
Kookmin UniversityIntelligence and Interaction Lab
LEE, Sang Hun
Education 1993 PhD Seoul National University, Mech. Eng. 1988 MS Seoul National University, Mech. Eng. 1993 BS Seoul National University, Mech. Eng.
Experience 1996 ~ Present Kookmin University, Professor 1999 Univ. of Washington, Visiting Scholar 1996 Advanced Engineering Institute, Researcher 1993 ~ 1995 SindoRicoh Co., Researcher
Research Interest Intelligent Human-Vehicle Systems and Interaction CAD and Human CAD for Automotive Design and Manufacturing Human-Machine Interaction (HMI) and HCI Machine Learning, Computer Graphics
Home Page Lab:http://ii.kookmin.ac.kr or http://cad.kookmin.ac.kr
Kookmin UniversityIntelligence and Interaction Lab
Facilities
Softwares Driving Simulation SW
PreScan, OpenDS, UC-win/Road
CAD/CAM CATIA, NX, Tecnomatics,
RapidForm Human CAD/CAE
AnyBody, Jack
Hardwares Driving Simulator
Kia K7
Driving Simulator
CATIA AnyBody
UC-win/Road
Kookmin UniversityIntelligence and Interaction Lab
Research Areas
Human-Vehicle Systems and Interaction Intelligent Driver-Vehicle Systems and User Interface Human-Vehicle Interaction Design and Engineering Applications of Machine Learning Technology
Digital Human Modeling and Simulation
Intelligence and Interaction Lab Kookmin University
Driving State Driver State Warning Device Control Autonomous Driving
Driver Intent Inference using Machine Learning Algorithms (1/4)
Background• Accidents due to carelessness, drowsiness and inexperienced
driving of the driver• Discrepancies between the various driving conditions, driver
intention and advanced driver assistance systems.• Increase in demand for a wide range of human-oriented
context-aware services.
Objective• Development of an active driver assistance system to respond
appropriately through a prediction of driver intention and traffic condition.
Approach• Generating predictive models for driving condition and driving
conditions using machine learning
Machine Learning
Vehicle
Driver
Intelligence and Interaction Lab Kookmin University
Essential Features
DriverHead Motion
Pupil Movement
Vehicle
Throttle Position
RPM
Speed
Longitudinal Acceleration
Steering Angle
Steering Velocity
Lateral Speed
Lateral Acceleration
Yaw Rate
Heading
Road
Time to Cross LaneLine
Offset from lanecenter
Surrounding Vehicle
Distance
Relative Speed
Time to Collision
ETC ː
DataAcquisition
Driver Intent Inference using Machine Learning Algorithms (2/4)
Kookmin UniversityIntelligence and Interaction Lab
Machine Learning Tool
Machine Learning Algorithm
Neural network
Support Vector Machine
Bayesian Network
···
Data Analysis, Model Design Pattern Recognition
Autonomous Driving
Driving State (Lane Change, Stop, Go, Turn, …)
Device Control
Driver StateWarning
Driver Intent Inference using Machine Learning Algorithms (3/4)
Intelligence and Interaction Lab Kookmin University
DisadvantagesIt’s hard to acquire the information of gaze
area. Too expensive
It’s hard to sustain the zero point during driving.
Essential Features
Driver
Head Motion
Pupil Movement(Stare area information)
Vehicle
Throttle Position
RPM
Speed
Longitudinal Acceleration
Steering Angle
Steering Velocity
Lateral Speed
Lateral Acceleration
Yaw Rate
Heading
Road
Time to Cross LaneLine
Offset from lanecenter
Surrounding Vehicle
Distance
Relative Speed
Time to Collision
ETC ː
SMI ETG DikablisKinect & Remodeled Webcam
OpenDS & Tobii & Faceshift & Structure sensor
Driver Intent Inference using Machine Learning Algorithms (4/4)
Intelligence and Interaction Lab Kookmin University
HMI for Multi-levels of Autonomous Vehicle(1/3)
Background• Spread autonomous vehicle due to various advanced driver
assistance systems(ADAS)• Driver maneuverability difficulties in the absence of a
unified HMI for autonomous vehicle (by SBD research 2013)
• Interface complexity, disconnection Driver is experiencing various confusion (by SBD research 2013)
Objective• To develop new driver interfaces for an autonomous vehicle
for effective interaction between intelligent system and driver.
Approach• Propose an driver-centric interface design approach• Perform driver-in-the-loop experiments
Operation degrees, Mode confusion when using systems, etc.
Kookmin UniversityIntelligence and Interaction Lab
Driver interface design & engineering• Design driver-centric interfaces• Develop mode transition chart consisting of various ADAS• Apply formal verification approach
HMI for Multi-levels of Autonomous Vehicle(2/3)
Kookmin UniversityIntelligence and Interaction Lab
Driver-in-the-loop experiment• Driver situation awareness, mode confusion evaluation• Design experimental roads and scenario using PreScan and
Matlab&Simulink PreScan: supporting development of driver assistance systems
and intelligent vehicle systems by providing an emulation function of various sensors (using Matlab & Simulink)
Absence: Eye track, Sound, Slow simulation
Dramoni (http://www.trywin.co.jp/item_detail/item_000016.html)
PreScan
Matlab & Simulink
HMI for Multi-levels of Autonomous Vehicle(3/3)
Kookmin UniversityIntelligence and Interaction Lab
Wheel Gesture Interaction Design
세부기능조절(ex :볼륨증가)
초기화면Back기능(ex :OFF)
초기 화면
메뉴 선택(ex :라디오)
[ 휠제스처 조작프로세스 ]
Kookmin UniversityIntelligence and Interaction Lab
Human-Vehicle Interaction Design & Analysis on Virtual Human-Product-Environment System
Kookmin UniversityIntelligence and Interaction Lab
Virtual Driver Based on ACT-R
Extension of ACT-R Visual and Motor Modules Comparison of Experimental and Simulation Results
Human Experiment ACT-R Simulation
Validation of Virtual Driver Development of Virtual Driver
Kookmin UniversityIntelligence and Interaction Lab
IntegratedHuman CAD
System
CAEMADYMO
(Crash Safety)
CAELifeMOD
(Biomechanics)
DigitalHuman
DigitalProduct
Parametric Human Modeling
Motion Simulation
Ingress/Egress Analysis
Integrated Human CAD/CAE System
Kookmin UniversityIntelligence and Interaction Lab
Human-centered Virtual Design and Analysis Process
Virtual Prototype
Motion & EMG Data Capture
Motion Reconstruction Biomechanical
Analysis
Ergonomic Evaluation
DB
Digital Human Generation
Motion Generation
• Motion• Force• EMG
Kookmin UniversityIntelligence and Interaction Lab
Human CAD
18
SimulationDMU of Vehicle + Digital Human
DMU of Vehicle
Tasks
Digital Manikin
-Sit-Hold Steering Wheel-Depress Pedal-Etc.
Reach/Motion Analysis
Vision Analysis
Comfort Assessment
Package Layout
Kookmin UniversityIntelligence and Interaction Lab
Virtual Integrated Development Environment (VIDE) for Green Car Parts
Kookmin UniversityIntelligence and Interaction Lab
Projects (selected)
운전자 상태와 주행 환경 분석을 통해 안전과 편의를 제공하는 Connectivity 기반의 개인맞춤 지능형 통합 Cockpit 모듈 개발 (지식경제부, 2013 ~ 2014)
제품과 인체의 통합 모델을 바탕으로 한 스포츠용품의 가상 설계 및 시험 프레임웍의 개발 (한국연구재단, 2013 ~ 현재)
근신경학적 퇴행 지연을 위한 디지털 인체 모델 기반 기반 통합 운동 시스템의 가상 설계 및시험 프레임웍 개발 연구 (2013 ~ 현재)
감성기반 지능형 자동차 인터랙션에 대한 공학-디자인 융합 연구 (한국연구재단, 2013) 인간친화적인 차량설계를 위한 인체동작 시뮬레이션 및 안락도 평가기술의 개발 (산학협동
재단, 2011~2012 ) 그린카 부품 상용화지원을 위한 가상개발환경(VIDE) 개발 (지식경제부, 2010~2012) 지식기반 3차원 금형설계 자동화 시스템 개발 (서울시 산학연, 2009~2011) 인간중심설계를 위한 제품 및 인체의 통합 CAD/CAE 시스템의 개발 (한국과학재단,
2008~2011) 지식기반 설계기술 개발 (GM대우 자동차, 2006~2008) 지능형 금형설계 시스템 (산업자원부, 2006~2007) 충돌 안전 해석을 위한 솔리드 중립면 생성 프로그램 (현대모비스, 2006~2007) 인간 친화적 미래형 자동차를 위한 인체 모델링과 시뮬레이션 시스템 개발 (교육인적자원부
/산업자원부/노동부, 2005~2007)