15|51 - Intergo › public › Publicaties › 9 › bestand... · Elements indicate minimum rather...

transcript

15|51 • 01-09-16 • 1

Program

14:00 Richard van der Weide, Intergo 14:15 Karel Brookhuis, Universiteit Groningen 14:45 Sarah Wiseman, University of London 15:15 Break 15:45 Chris Jansen, Universiteit Utrecht 16:15 Marc Woods, Mindhacker

Stella Donker, Universiteit Utrecht 17:00 Drinks

15|51 • 01-09-16 • 2

Marit Wilms

Richard vd Weide

Melcher Zeilstra

Diana Went

Gert-Jan Kamps

Kirsten Schreibers Fenneke Blommers

David de Bruijn

Alfred v Wincoop Henk Frieling

Colete Weeda

Deconducteurvandetoekomst

Program

15|51 • 01-09-16 • 14

Karel Brookhuis (Traffic Psychology)

Traffic Psychology Group, Dept. Neuropsychology

Faculty Behavioural & Social Sciences

University of Groningen (Delft University of Technology) Colleagues: Dick de Waard, Janet Veldstra, Anselm Fürmaier, Arjan Stuiver

Human Factors in traffic Rijksuniversiteit Groningen

Faculty of Behavioural & Social Sciences Department of Psychology

Experimental & Work Psychology

Traffic in 1900

1900 Seat belt Patent by Gustave-Desire Levau

The rationale of Traffic Psychology

Accidents at least 95% “Human Factor” The driver makes errors. is not alert, distracted, tired, etc.

Development of fatal accidents in traffic in the Netherlands (SWOV)

Jaar1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 20000

1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000

05 September 2016 5

Research on in-vehicle ICT 25 years of driver support

• European projects, off 1989 • GIDS (Generic Intelligent Driver Support System) • AWAKE (Drowsiness warning)

• Dutch national initiatives (Ministry) • ISA (Intelligent Speed Assistance) • Belonitor (Rewarding good i.s.o. Punishing bad) • PAYD (Pay As You Drive, with Insurance cies)

• National Research Council (NWO) programs • PITA (Personal Intelligent Travel Assistant) • Amici (Advanced Multi-agent Information and Control for

Integrated traffic networks) • MDPIT (Multidisciplinary pricing policy)

05 September 2016 6

25 years of research on effects of driver support & information supply, EU start: Generic Intelligent Driver Support (European Framework, DRIVE 1989 - 1991)

• Philips (CARIN) & Bosch navigation systems • RDS – TMC (Radio Data System – Traffic

Message Channel)

• Simulator studies (VSC/TRC, Groningen) • Instrumented Vehicle (TNO, Soesterberg)

05 September 2016 7

GIDS: (early, 1990) Carin navigation system

05 September 2016 8

( GIDS in 1989 … simulator 2010)

05 September 2016 9

Results EU studies, GIDS … AutoVeh • Problems: attention, distraction, alertness • Accident likelihood:

• 2 a 2.5 second rule (within car) • 4 second rule (outside car) • 15 second rule (total processing time)

• Combination auditory and visual information • Improved route guidance systems

• GIDS book, Michon et al., T&F 1993 • Guidelines travel / traffic information

• HASTE Special Issue TR / Part F • In-Safety (Book, 2010) • Special Issue TR/F (History Traffic Psychology, 2015)

05 September 2016 10

Intelligent Speed Assistant ISA (1995, in advisory version)

Results of Feedback

Compound opinions on ISA

-3 -2 -1 0 1 2

Pleasurable OLD

Practical OLD

Pleasurable YOUNG

Practical YOUNG

BeforeAfter

ISA effects: 36% less injury accidents, up to 59% less fatal accidents

Longitudinal effects: wears off, but not completely

12 maart 2008 Symposium Synergie in Mobiliteit, Universiteit Twente

Ministry (NL) funded project

New concept: reward i.s.o. punish

• distance: 1.3 s

• local speed limit

• 62 lease-car drivers

• 4 weeks baseline

• 16 weeks test

• 4 weeks baseline

• subjective and objective measurements

Belonitor – Rewarding safe driving behaviour

Earned credits

Bad behaviour

Good behaviour

Belonitor - Conclusions General Attitude: 68% positive about system, 75% improved behaviour

Traffic Safety • Estimated reduction lethal & severely injured 15% • Estimated reduction injured 9% Efficiency and Environmental • reduction accidents 10% – traffic jams 1.2%

• reduction fule use 5.5 %

Bron: MinVenW, 2006

USA 2013: statistics indicates a 4% to 5% increase in accident involvement

Side effects of Lane Departure Warning

2006: RouteLint HSM (Railway-Project Het Spoor Meester)

Attention for action → Distraction

Multitasking ↔ Distraction

Thanks to Peter Hancock

27 Driver distraction and inattention in its various forms is thought to play a role in 20-30% of all road

crashes (Wang, Knipling & Goodman, 1996).

Future ?? (Google stopped !!) =>ICT in traffic

http://www.youtube.com/watch?v=V8ofTlynWPo&feature=player_detailpage

So, no more human drivers ? Automated Vehicle !

Opleiding Psychologie/Faculteit gedrags- en

maatschappijwetenschappen

Summary of Levels of Driving Automation for On-Road Vehicles This table summarizes SAE International’s levels of driving automation for on-road vehicles. Information Report J3016 provides full definitions for these levels and for the italicized terms used therein. The levels are descriptive rather than no rmative and technical rather than legal. Elements indicate minimum rather than maximum capabilities for each level.

“System" refers to the driver assistance system, combination of driver assistance systems, or automated driving system, as appropriate.

The table also shows how SAE’s levels definitively correspond to those developed by the Germany Federal Highway Research Institute (BASt) and approximately correspond to those described by the US National Highway Traffic Safety Administration (NHTSA) in its “Pre liminary Statement of Policy Concerning Automated Vehicles” of May 30, 2013.

Name Narrative definition

Execution of steering and acceleration/ deceleration

Monitoring of driving

environment

Fallback performance of dynamic driving task

System capability (driving modes) B

Human driver monitors the driving environment

0 No Automation

the full-time performance by the human driver of all aspects of the dynamic driving task, even when enhanced by warning or intervention systems Human driver Human driver Human driver n/a

1 Driver Assistance

the driving mode-specific execution by a driver assistance system of either steering or acceleration/decelerati on using information about the driving environment and with the expectation that the human driver perform all remaining aspects of the

dynamic driving task

Human driver and system Human driver Human driver

Some driving modes A

2 Partial Automation

the driving mode-specific execution by one or more driver assistance systems of both steering and acceleration/deceleration using information about the driving environment

and with the expectation that the human driver perform all remaining aspects of the dynamic driving task

System Human driver Human driverSome driving modes P

Automated driving system (“system”) monitors the driving environment

3 Conditional Automation

the driving mode-specific performance by an automated driving system of all aspects of the dynamic driving task with the expectation that the human driver will respond

appropriately to a request to intervene System System Human driver

Some driving modes H

4 High Automation

the driving mode-specific performance by an automated driving system of all aspects of the dynamic driving task, even if a human driver does not respond appropriately to a

request to intervene System System System

Some driving modes

5 Full Automation

the full-time performance by an automated driving system of all aspects of the dynamic driving task under all roadway and environmental conditions that can be managed by a

human driver System System System All driving

Source: SAE Standard J3016 Report

Summary of Levels of Driving Automation for On-road Vehicles

› Simulator experiment in Groningen • Drivers of Hermes and Arriva companies • Trajectory Eindhoven CS to Eindhoven Airport

Automating Public Transport SAE level 2

Phileas (APTS bv)

› HOV in Eindhoven • Hybrid system and control: (half)automatic and normal • Phileas (Automatic Public Transport System)

› Simulator experiment in Groningen, • Bus drivers of Hermes and Arriva • Track Eindhoven CS to Eindhoven Airport

Automating Public Transport 1996 (Hoogwaardig Openbaar Vervoer)

Incident 1: vehicle blocks the road

o Braking in time

o Bus back to non- or semi-automatic

Incident 2: a cyclist runs red light

o 1st time: 72% okay, but 28 % not !!!

o 2nd time: 100% okay !

Incident 1: vehicle blocks the road

o Braking in time

o Bus back to non- or semi-automatic

Experiment in driving simulator SAE level 3

Autonomous driving

Emergency situation on automated highway

Reactions to the unexpected event

Reactions Time slot Proportion participants No reaction - f 50 % Braked Late > 14 s 15 % Braked Not fast 9-14 s 30 % Braked Early < 9 s 5 %

Attitude towards fully automated system Before After › Usefulness 0.53 0.38 › Satisfying –0.31 –0.36

The road towards fully autonomous vehicles: via SAE levels 2 and 3 ??

Perhaps Chris Janssen will tell us

Mental Model, evidence form another field…. (thanks to Dietrich Manzey) According to Malinge (2011), the reduction of the annual accident rate of fourth generation airlines has stagnated, notwithstanding the inherent effectiveness. Although automated systems have evolved to reduce pilot workload, recent concerns exist about their safety effectiveness. Automation introduces a paradox: providing crews with necessary operational assistance simultaneously dissociates the crew from those operations. Unfortunately, this shift in pilot tasking exhibits itself in many forms of adverse crew behavior such as automation induced complacency (Manzey, Reichenbach & Onnasch, 2012), automation bias (Mosier, Skitka, Heers & Burdick, 1998), decision making errors (Orasanu, Martin & Davison, 1998), lack of (procedural and declarative) system knowledge and/or manual control skills (Potter et al., 2012), which are all in turn aggravated by overconfidence (Wood, 2004) and fatigue (Caldwell, 2012). These behaviors may contribute to the loss of situational awareness (SA).

18 March 2016 47 www.its.leeds.ac.uk

Survey: Safety and Priority?

So, automation… ?

When, how, where, who ?

Kyriakidis et al. “A Human Factors perspective on Automated Driving ” (i.e. European perspective), in preparation

k.a.brookhuis@rug.nl

Program

15|51 • 01-09-16 • 64

Everything you always wanted to know about

numbers, but were afraid to ask

Dr Sarah WisemanGoldsmiths, University of London

Who am I? •  Currently: !

•  Research and teaching fellow at Goldsmiths!

•  Public speaker on HCI and HF!

•  In the past:!

•  HCI PhD looking at number entry interfaces!

Who am I? •  Currently: !

•  Research and teaching fellow at Goldsmiths!

•  Public speaker on HCI and HF !

•  In the past:!

•  HCI PhD looking at number entry interfaces!

•  Daily Mail accredited “scientist” !

Why do you care about numbers?

Newtons Pounds per square inch

Number entry interfaces in the hospital

Electronic patient !records!

Infusion pump !devices!

Radiography / X-ray !equipment!

Why are number entry interfaces designed

like that?

Telephone Calculator

Expectations

all day long

But why are they laid out in a grid like

Reality

“offered certain engineering advantages”

Desires

Can we improve the standard number entry interface?

Improving text entry

URL email

Ok before we improve anything, what’s the most

common number?

Benford’s Law

hUp://www.thecleverest.com/benfords_law.html

When does Benford apply? “Naturally occurring cumulative numbers”

Number of twitter followers Length of rivers

Electricity bills First 652066 Fibonacci Numbers

When does Benford apply?

When does it not apply?

“Naturally occurring cumulative numbers”

Lottery numbers

Fraudulent invoices/expenses

Lottery numbers

Drug dosages in hospitals

Numbers used in the hospital

100, 5.5, 300, 1000, 750, 33.33, 999, 12.25, 150, 200, 40, 600…

Numbers used in the hospital

Redesigning the interface

Significantly faster

No increase in error rate

Redesigning the interface

Significantly faster

No increase in error rate

Is the interface design all there is to think about?

(You are a very smart audience)

Number entry is a 3-step process

1. Read the number

2. Memorise the number

3. Type the number

Number entry is a 3-step process

1. Read the number

2. Memorise the number

3. Type the number

Sarah Wiseman www.swiseman.co.uk

@oopsohno s.wiseman@gold.ac.uk

Program

15|51 • 01-09-16 • 154

9/5/16&

Automation and HCI: A case study of (semi-) autonomous car

Chris Janssen c.p.janssen@uu.nl www.cpjanssen.nl

H2020-MSCA-IF-2015, 705010, 'Detect and React'

http://jalopnik.com/teslas-autopilot-system-is-awesome-and-creepy-and-a-sig-1736573089

9/5/16&

(semi-) Autonomous cars: Challenge? SAE level

Name Description

Human driver monitors driving environment

0 No Automation

1 Driver Assistance

2 Partial Automation

Automatic driving system monitors driving system

3 Conditional Automation

4 High Automation

5 Full Automation

SAE International (2014)

Designing for automated systems

Linda Boyle U Washington

Andrew Kun U New Hampshire

Lewis Chuang Max Planck

Wendy Ju Stanford

Initiated at Dagstuhl 16262

Driving requires varied behavior..

Strategic

Maneuvering

Control

After Michon (1985)

Lane position detection

Traffic detection

Change lanes

Driving requires varied behavior..

Lane position detection

Traffic detection

Change lanes ✔

9/5/16&

Non-automated car

Change lanes Traffic detection Lane detection

Fully automated car (e.g., “ideal” google car)

Partial automated car (e.g. “Tesla”)

Potential dangers / issues •  Transitions:

-  How to inform human about transition? -  How does driver know which state they

are in? -  How to inform and enforce a shift in

responsibility?

9/5/16&

Broader context: automation How do we.. •  Communicate change? •  Design for shared responsibility? •  Assess system state? •  Assess human state?

Example studies: asses human state

1.  ‘Looking at the eyes’

2.  ‘Sensing in the brain’

1. ‘Looking at the eyes’

Hidde van der Meulen MSc AI Utrecht

Andrew Kun U New Hampshire

Van der Meulen, Kun, Janssen (2016) Proceedings ACM AutoUI

Set-up

9/5/16&

Eye-gaze

Figure 4. Top-figure: The percentage of looking at the road center (PRC) as a function of time. Bottom-figure: the average driving

speed over time. In both figures, the vertical dashed line shows the moment in time where both conditions have the same speed, with the standard deviation displayed as vertical grey bars. The error bars and ribbon show the standard error values of each

metric.

We used a 24 (time after take-over: buckets of 5s) x 2 (situation: parking or autonomous) within-subjects ANOVA to determine the effect of time and condition on the PRC. The time after take-over had a significant influence on the PRC with F(1, 23) = 3.73, p < .001, there was no significant effect for driving type with F(1, 1) = 0.93, p = .542 and no significant interaction effect between time after take-over and the type of driving with F(1, 23) = 0.55, p = .954. As Figure 4 shows, the PRC gradually increases in the first 15 to 25 seconds and then stabilizes

Percent dwell time Figure 5 plots the PDT score for the autonomous driving condition (dark grey bars) and the parking condition (light grey bars) for the 100 seconds before taking over control (left two bars) and 2 minutes after taking over control (right two bars). A 2 (timing: before, after take-over) x 2 (situation: parking or autonomous) within-subjects ANOVA revealed that there was a main effect of timing, such that drivers looked more at the road after taking over, F(1, 15) = 78.75, p < .001. There was also a main effect of driving situation, F(1, 15) = 7.39, p = .016. This was influenced by an interaction effect, F(1, 15) = 7.48, p = .015.

As Figure 5 shows, the interaction was such that before the take-over, the participants looked almost twice as often at the road in the autonomous driving condition (M = 48%, SD = 9.4%) compared to the parking condition (M = 27%, SD = 6.22%), whereas after take-over both groups spent roughly a

similar amount of time gazing at the road (in both conditions M = 99%, SD = 0.5%).

Figure 5. Percentage of time spent looking at the road before

and after the take-over for both autonomous driving and parking conditions.

2. ‘Sensing the brain’

Remo vd Heiden Promovendus

Leon Kenemans Chantal Merkx Rijkswaterstaat

Stella Donker

H2020-MSCA-IF-2015, 705010, 'Detect and React'

Data: BSc thesis Lotte Hardeman & Keri Mans

9/5/16&

0 500 1000 1500

P3a at FCz (Novel - Standard)

Time (ms)

StationaryAutonomousDriving

(Active)

Preliminary results Do not cite

0 500 1000 1500

P3a at FCz (Novel - Standard)

Time (ms)

StationaryAutonomousDriving

(Passive)

Preliminary results Do not cite

Broader context: automation How do we.. •  Communicate change? •  Design for shared responsibility? •  Assess system state? •  Assess human state?

Program

15|51 • 01-09-16 • 188

Mindhacking:factorficFon?

StellaDonkers.f.donker@uu.nl

Why does “mindhacking” work?

Mindhacking in the “real” world

Why does mindhacking work?

PercepFonislearned:BrainislazyPaUernrecogniFonEfficienttojumptoconclusions

DirecFngaUenFon

In/uitcheckenAmstelstaFon

DirecFngaUenFonin the “real” world

IncheckenamstelstaFon

DirecFngaUenFonin the “real” world

Failureofawareness:theinformaFonistherebut(some)peopledonotseeit

Failure of awareness

Lookedbutfailedtosee(SMIDSY)

“Sorry mate, didn’t see you”

TheinformaFonisthere,butsFll(some)peopledonotseeit

Mindhackingisafact

Mindhacking:FactorFicFon?

istheresultofthemagnificantpowerofthebrain.

ThefactthatamindhackercanmakeficFonfeelreal,

Program

15|51 • 01-09-16 • 206

15|51 - Intergo › public › Publicaties › 9 › bestand... · Elements indicate minimum rather...

Documents