TRB Conference on Automated Driving Fundamentals and ...
21.07.2015
TRB Conference on Automated Driving
Fundamentals and Experiments on handing over the Driving Task
21st July 2015 Univ.-Prof. Dr.-Ing. Lutz Eckstein Dipl.-phil. Johanna Josten
Institute for Automotive Engineering (ika), RWTH Aachen University Prof. Dr. Lutz Eckstein
Slide No. 1
© ika 2015 · All rights reserved
21.07.2015
Introduction ika & fka Facilities and Infrastructure
6
7 3
4
5
1 fka-building & German Design Studio Aachen 2 ika-building
2
est track t ac and a d 3 Test driving simulator 4 Test bench building 5 Centre for metal construction
1
Aachen Prof. Dr. Lutz Eckstein
Slide No. 2
21.07.2015
6 Aldenhoven Testing Center (ATC) 7 High-fidelity driving simulator © ika 2015 · All rights reserved
1
21.07.2015
Fundamentals Levels of automated driving – Hypothesis: Automation Gap
Driving Task
Human
Level of Automation
Navigation Choice of route Driver controls
Guidance Choice of course and velocity
Stabilization Steering wheel and pedal operation
Driver monitors …
Driver delegates
2
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
driverless
Highly automated
informed Level 0
Driver needs to be sufficiently motivated to monitor State-of-the art = driver needs to take action from time to time
Level 3
Level 4
Level 5
Traffic jjam Assist
1
Driver needs to be sufficiently stimulated to take over Suitable cues and timing?
2 Fail-safe system
Fail-silent System Prof. Dr. Lutz Eckstein
Slide No. 3
hands-on vs. hands-off short vs. long system activity before TOR Autobahn Pilot, non-driving task on iPad div. Modalities of TOR TOR piano vs. forte © ika 2015 · All rights reserved
21.07.2015
Fundamentals Level 3: Handing over the driving task Driver Perception Information Processing
Non-driving Non driving Tasks
Vehicle Control Elements
Action
Logic
Displays
System-based S t b d Functions
T ti Texting, Phoning, Ph i Browsing, B i Relaxing, …
Navigation Choice of route
Vehicle behaviour
Guidance Choice of course and velocity
Automated Driving
Stabilization Steering St i wheel h l and d pedal operation
Sensors
Function
Actuators
Driving Task
Vehicle performs the driving task in defined situations (e.g. v < 60 km/h on highways) Prof. Dr. Lutz Eckstein
Slide No. 4
21.07.2015
© ika 2015 · All rights reserved
2
21.07.2015
Fundamentals Level 3: Handing over the driving task Glance
Contact
Driver t2
t1
Perception Information Processing
Non-driving Non driving Tasks
start action
end of correction
t3
t4
Vehicle
Control Elements
Action
Logic
Displays
System-based S t b d Functions
T ti Texting, Phoning, Ph i Browsing, B i Relaxing, …
Navigation Vehicle behaviour Transition from automatedAssisted to assisted driving Driving Automated Driving
act
decide
evaluate
recognize
Guidance Choice of course and velocity
perceive
Choice of route
Stabilization Steering St i wheel h l and d pedal operation
Driving Task
Sensors
Function
Actuators
Information
Need for valid HMI-criteria to design the hand-over in a safe way => HMI-guideline Prof. Dr. Lutz Eckstein
Slide No. 5
© ika 2015 · All rights reserved
21.07.2015
Human
Experiment 1 Research question and experimental design
Le ve l o f Au to m ati on
Driver controls
Driver monitors …
1
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
Driver delegates driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Goal: Analysis of drivers’ handling performance in short-term take-over situations after partial automated driving
Comparison between handling performance after manual and after partial-automated driving
Influence of hand posture and duration of system use on handling handling-performance performance in take-over take over situations
Two studies in motion based driving simulator:
Validation of results on test-track:
1. Variation of hand posture and duration of partly automated driving
Variation of hand posture Comparison with handling-performance as measured in driving simulator
2. Investigation of selected influencing factors, e.g. non-driving related tasks
Prof. Dr. Lutz Eckstein
Slide No. 6
21.07.2015
© ika 2015 · All rights reserved
3
21.07.2015
Human
Experiment 1 Driving Simulator Study 1: First Results
Le ve l o f Au to m ati on
Driver monitors
Driver controls
1
…
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
Driver delegates driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Preliminary results, further subjects under test
Prof. Dr. Lutz Eckstein
Slide No. 7
© ika 2015 · All rights reserved
21.07.2015
Human
Experiment 2: highly automated traffic jam pilot Experimental design and key findings
Le ve l o f Au t o m at i on
Driver controls
Driver monitors …
assisted
Partly automated
Level 1
Level 2
System
Driver delegates
2
fully automated
driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Goal: Influence of HMI-Modality and Intensity on take over process, timing and quality
HMI variations: Speech, seat vibration, visual (LED running lights)
Performance of younger (25 – 35) vs. vs older drivers (50 – 70)
Function was switched off on motorway in a curve immediately after take-over request
Prof. Dr. Lutz Eckstein
Slide No. 8
21.07.2015
© ika 2015 · All rights reserved
4
21.07.2015
Human
Experiment 2: highly automated traffic jam pilot Key findings
Le ve l o f Au t o m at i on
Driver controls
Driver monitors …
assisted
Partly automated
Level 1
Level 2
System
Driver delegates
2
fully automated
driverless
Highly automated
informed Level 0
Acoustic take-over signal 3000
Baseline
High Intensity
Low Intensity
Baseline
High Intensity
Low Intensity
1500
1000
2000
1500
1000
Low Intensity
1500
1000
500
0
t1 t2 tw tk Mean reaction times with standard errors
High Intensity
2000
500
500
Baseline
2500 Reaction time (in ms)
2000
Level 5
Visual take-over signal 3000
2500 Reaction time (in ms)
Reaction time (in ms)
Level 4
Haptic take-over signal 3000
2500
0
Level 3
0
ttw ttk 1 2 Mean reaction times with standard errors
t1 ttk tw 2 Mean reaction times with standard errors
2 to 2.5 seconds until driver has contact with controls (t2), action follows almost immediately (t3) Duration until end of action (t4) depends on subject, situation and HMI and may take up to 10 seconds
More research on modality x intensity needed Benefit of ADAS after take-over to be investigated
Multimodal, multistage HMI-concept seems advisable Prof. Dr. Lutz Eckstein
Slide No. 9
© ika 2015 · All rights reserved
21.07.2015
Visualization of trade-offs using the 4-level-model by Eckstein Example and Outlook Acceptance by different societal groups
DRIVER
Societal Level
VEHICLE Rules on driver behavior Which driver activities are allowed during automated driving
Legal Level
Human Factors Level
VEHICLE Driver - Vehicle Interaction Safe transition = f (secondary tasks) Transition time = f (HMI, functionality)
VEHICLE HMI - Concept Information presentation: Modalities Logics & Timing: escalation and timing
Technical Level Functional transition Fade-out of automated driving Driver assistance after TOR
S O C I E TEYN V I R O N M E N T
DRIVER ENVIRONMENT
DRIVER ENVIRONMENT Driver Monitoring Actual use of system TOR = f (driver state)
DRIVER
VEHICLE
ENVIRONMENT
TECHNICAL STANDARDS Prof. Dr. Lutz Eckstein
Slide No. 10
21.07.2015
© ika 2015 · All rights reserved
5
21.07.2015
Summary
Driver
Automated driving constitutes are multi-level challenge, ranging from a technical level via a human factors level, a legal level to a societal level.
Perception Information Processing
Non-driving Tasks
Vehicle Action
Control Elements
Texting, Phoning, Browsing, Relaxing, …
Logic
Displays
System-based Functions
Navigation Choice of route
Vehi beha
Guidance Choice of course and velocity
Automated Driving
Stabilization
The human factors level plays a central role role, since automated driving – like every technology – has limits requiring interaction between vehicles and human beings.
Interaction is not only required in case of handing over the driving task to the driver, but also during automated driving.
Challenges require new concepts of interdisciplinary and international research:
linking automotive engineering to human factors, computer science, psychology,
Creating a high-fidelity driving simulator, depicting driving on motorways 1:1
Establishing a R&D facility in Silicon Valley, part of ProspectSV in San José
Prof. Dr. Lutz Eckstein
Steering wheel and pedal operation
Sensors
Function
Actuators
Driving Task
Slide No. 11
21.07.2015
© ika 2015 · All rights reserved
Slide No. 12
21.07.2015
© ika 2015 · All rights reserved
Contact
Univ.-Prof. Dr.-Ing. Lutz Eckstein
Institute for Automotive Engineering (ika) RWTH Aachen University Steinbachstr. 7 52074 Aachen Germanyy
Phone Fax
+49 241 – 80 25600 +49 241 – 80 22147
Email [email protected] Internet www.ika.rwth-aachen.de Prof. Dr. Lutz Eckstein
6
TRB Conference on Automated Driving
Fundamentals and Experiments on handing over the Driving Task
21st July 2015 Univ.-Prof. Dr.-Ing. Lutz Eckstein Dipl.-phil. Johanna Josten
Institute for Automotive Engineering (ika), RWTH Aachen University Prof. Dr. Lutz Eckstein
Slide No. 1
© ika 2015 · All rights reserved
21.07.2015
Introduction ika & fka Facilities and Infrastructure
6
7 3
4
5
1 fka-building & German Design Studio Aachen 2 ika-building
2
est track t ac and a d 3 Test driving simulator 4 Test bench building 5 Centre for metal construction
1
Aachen Prof. Dr. Lutz Eckstein
Slide No. 2
21.07.2015
6 Aldenhoven Testing Center (ATC) 7 High-fidelity driving simulator © ika 2015 · All rights reserved
1
21.07.2015
Fundamentals Levels of automated driving – Hypothesis: Automation Gap
Driving Task
Human
Level of Automation
Navigation Choice of route Driver controls
Guidance Choice of course and velocity
Stabilization Steering wheel and pedal operation
Driver monitors …
Driver delegates
2
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
driverless
Highly automated
informed Level 0
Driver needs to be sufficiently motivated to monitor State-of-the art = driver needs to take action from time to time
Level 3
Level 4
Level 5
Traffic jjam Assist
1
Driver needs to be sufficiently stimulated to take over Suitable cues and timing?
2 Fail-safe system
Fail-silent System Prof. Dr. Lutz Eckstein
Slide No. 3
hands-on vs. hands-off short vs. long system activity before TOR Autobahn Pilot, non-driving task on iPad div. Modalities of TOR TOR piano vs. forte © ika 2015 · All rights reserved
21.07.2015
Fundamentals Level 3: Handing over the driving task Driver Perception Information Processing
Non-driving Non driving Tasks
Vehicle Control Elements
Action
Logic
Displays
System-based S t b d Functions
T ti Texting, Phoning, Ph i Browsing, B i Relaxing, …
Navigation Choice of route
Vehicle behaviour
Guidance Choice of course and velocity
Automated Driving
Stabilization Steering St i wheel h l and d pedal operation
Sensors
Function
Actuators
Driving Task
Vehicle performs the driving task in defined situations (e.g. v < 60 km/h on highways) Prof. Dr. Lutz Eckstein
Slide No. 4
21.07.2015
© ika 2015 · All rights reserved
2
21.07.2015
Fundamentals Level 3: Handing over the driving task Glance
Contact
Driver t2
t1
Perception Information Processing
Non-driving Non driving Tasks
start action
end of correction
t3
t4
Vehicle
Control Elements
Action
Logic
Displays
System-based S t b d Functions
T ti Texting, Phoning, Ph i Browsing, B i Relaxing, …
Navigation Vehicle behaviour Transition from automatedAssisted to assisted driving Driving Automated Driving
act
decide
evaluate
recognize
Guidance Choice of course and velocity
perceive
Choice of route
Stabilization Steering St i wheel h l and d pedal operation
Driving Task
Sensors
Function
Actuators
Information
Need for valid HMI-criteria to design the hand-over in a safe way => HMI-guideline Prof. Dr. Lutz Eckstein
Slide No. 5
© ika 2015 · All rights reserved
21.07.2015
Human
Experiment 1 Research question and experimental design
Le ve l o f Au to m ati on
Driver controls
Driver monitors …
1
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
Driver delegates driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Goal: Analysis of drivers’ handling performance in short-term take-over situations after partial automated driving
Comparison between handling performance after manual and after partial-automated driving
Influence of hand posture and duration of system use on handling handling-performance performance in take-over take over situations
Two studies in motion based driving simulator:
Validation of results on test-track:
1. Variation of hand posture and duration of partly automated driving
Variation of hand posture Comparison with handling-performance as measured in driving simulator
2. Investigation of selected influencing factors, e.g. non-driving related tasks
Prof. Dr. Lutz Eckstein
Slide No. 6
21.07.2015
© ika 2015 · All rights reserved
3
21.07.2015
Human
Experiment 1 Driving Simulator Study 1: First Results
Le ve l o f Au to m ati on
Driver monitors
Driver controls
1
…
fully automated
1
assisted
Partly automated
Level 1
Level 2
System
Driver delegates driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Preliminary results, further subjects under test
Prof. Dr. Lutz Eckstein
Slide No. 7
© ika 2015 · All rights reserved
21.07.2015
Human
Experiment 2: highly automated traffic jam pilot Experimental design and key findings
Le ve l o f Au t o m at i on
Driver controls
Driver monitors …
assisted
Partly automated
Level 1
Level 2
System
Driver delegates
2
fully automated
driverless
Highly automated
informed Level 0
Level 3
Level 4
Level 5
Goal: Influence of HMI-Modality and Intensity on take over process, timing and quality
HMI variations: Speech, seat vibration, visual (LED running lights)
Performance of younger (25 – 35) vs. vs older drivers (50 – 70)
Function was switched off on motorway in a curve immediately after take-over request
Prof. Dr. Lutz Eckstein
Slide No. 8
21.07.2015
© ika 2015 · All rights reserved
4
21.07.2015
Human
Experiment 2: highly automated traffic jam pilot Key findings
Le ve l o f Au t o m at i on
Driver controls
Driver monitors …
assisted
Partly automated
Level 1
Level 2
System
Driver delegates
2
fully automated
driverless
Highly automated
informed Level 0
Acoustic take-over signal 3000
Baseline
High Intensity
Low Intensity
Baseline
High Intensity
Low Intensity
1500
1000
2000
1500
1000
Low Intensity
1500
1000
500
0
t1 t2 tw tk Mean reaction times with standard errors
High Intensity
2000
500
500
Baseline
2500 Reaction time (in ms)
2000
Level 5
Visual take-over signal 3000
2500 Reaction time (in ms)
Reaction time (in ms)
Level 4
Haptic take-over signal 3000
2500
0
Level 3
0
ttw ttk 1 2 Mean reaction times with standard errors
t1 ttk tw 2 Mean reaction times with standard errors
2 to 2.5 seconds until driver has contact with controls (t2), action follows almost immediately (t3) Duration until end of action (t4) depends on subject, situation and HMI and may take up to 10 seconds
More research on modality x intensity needed Benefit of ADAS after take-over to be investigated
Multimodal, multistage HMI-concept seems advisable Prof. Dr. Lutz Eckstein
Slide No. 9
© ika 2015 · All rights reserved
21.07.2015
Visualization of trade-offs using the 4-level-model by Eckstein Example and Outlook Acceptance by different societal groups
DRIVER
Societal Level
VEHICLE Rules on driver behavior Which driver activities are allowed during automated driving
Legal Level
Human Factors Level
VEHICLE Driver - Vehicle Interaction Safe transition = f (secondary tasks) Transition time = f (HMI, functionality)
VEHICLE HMI - Concept Information presentation: Modalities Logics & Timing: escalation and timing
Technical Level Functional transition Fade-out of automated driving Driver assistance after TOR
S O C I E TEYN V I R O N M E N T
DRIVER ENVIRONMENT
DRIVER ENVIRONMENT Driver Monitoring Actual use of system TOR = f (driver state)
DRIVER
VEHICLE
ENVIRONMENT
TECHNICAL STANDARDS Prof. Dr. Lutz Eckstein
Slide No. 10
21.07.2015
© ika 2015 · All rights reserved
5
21.07.2015
Summary
Driver
Automated driving constitutes are multi-level challenge, ranging from a technical level via a human factors level, a legal level to a societal level.
Perception Information Processing
Non-driving Tasks
Vehicle Action
Control Elements
Texting, Phoning, Browsing, Relaxing, …
Logic
Displays
System-based Functions
Navigation Choice of route
Vehi beha
Guidance Choice of course and velocity
Automated Driving
Stabilization
The human factors level plays a central role role, since automated driving – like every technology – has limits requiring interaction between vehicles and human beings.
Interaction is not only required in case of handing over the driving task to the driver, but also during automated driving.
Challenges require new concepts of interdisciplinary and international research:
linking automotive engineering to human factors, computer science, psychology,
Creating a high-fidelity driving simulator, depicting driving on motorways 1:1
Establishing a R&D facility in Silicon Valley, part of ProspectSV in San José
Prof. Dr. Lutz Eckstein
Steering wheel and pedal operation
Sensors
Function
Actuators
Driving Task
Slide No. 11
21.07.2015
© ika 2015 · All rights reserved
Slide No. 12
21.07.2015
© ika 2015 · All rights reserved
Contact
Univ.-Prof. Dr.-Ing. Lutz Eckstein
Institute for Automotive Engineering (ika) RWTH Aachen University Steinbachstr. 7 52074 Aachen Germanyy
Phone Fax
+49 241 – 80 25600 +49 241 – 80 22147
Email [email protected] Internet www.ika.rwth-aachen.de Prof. Dr. Lutz Eckstein
6
