PDF VERSION BY · PDF filePDF VERSION BY PDF VERSION BY PDF VERSION BY
Deliverable 10.4 Project Final Report€¦ · ADAS&ME (688900) D10.4 – Final project report...
Transcript of Deliverable 10.4 Project Final Report€¦ · ADAS&ME (688900) D10.4 – Final project report...
Adaptive ADAS to support incapacitated drivers Mitigate Effectively risks through tailor
made HMI under automation
This project has received funding from the European Union’s Horizon 2020
research and innovation programme under grant agreement No 688900
Deliverable 10.4 – Project Final Report
Deliverable Identity
Work Package No. WP10
Work Package Title Management
Activity No. A10.1
Activity Title Administrative and Overall Management
Dissemination level PU = Public
Main Author(s)
Anna Anund (VTI)
Lena Nilsson (VTI)
File Name ADASANDME_Deliverable_xx.x_dd-mm_y.doc
Online resource http://www.adasandme.com
Ref. Ares(2020)1270577 - 28/02/2020
ADAS&ME (688900) D10.4 – Final project report
February 2020 Page 2 of 68 Version 2.1
Version History
Date Version Comments
2020-02-16 1.0 First draft made by VTI with input from WP-leaders
and use case leaders
2020-02-19 1.1 Reviewed by Quality manager
2020-02-23 2.0 Revised by VTI
2020-02-26 2.1 Revised by CERTH
Authors (full list)
Anna Anund, VTI
Lena Nilsson, VTI
Andreas Absnër, Scania
Frederik Diedrich, Fraunhofer
Marc Figuls, RACC
James Jackson, IDIADA
Christoph Allig, Denso
Marcel Mathissen, Ford
Eleonora Meta, CTL
Stella Nikolaou, CERTH
Kevin Nguyen, Valeo
Davide Sette, Ducati
Sri Venkata Naga Phanindra Akula, TUC
Marc Wilbrink, DLR
Project Coordinator
Dr. Anna Anund
Research Director / Associate Professor
VTI - Olaus Magnus väg 35 / S-581 95 Linköping / Sweden
Tel: +46-13-20 40 00 / Direct: +46-13-204327 / Mobile: +46-709 218287
E-mail: [email protected]
Legal Disclaimer
The information in this document is provided “as is”, and no guarantee or warranty is given that the information
is fit for any particular purpose. The above referenced authors shall have no liability for damages of any kind
including without limitation direct, special, indirect, or consequential damages that may result from the use of
these materials subject to any liability which is mandatory due to applicable law.
The present document is a draft. The sole responsibility for the content of this publication lies with the authors. It
does not necessarily reflect the opinion of the European Union. Neither the INEA nor the European Commission
is responsible for any use that may be made of the information contained therein.
© 2016 by ADAS&ME Consortium
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 3 of 68 Version 2.1
Table of Contents
1 THE SCOPE OF ADAS&ME ...................................................................................................................... 11
1.1 OBJECTIVES ................................................................................................................................................ 13 1.1.1 Objectives of ADAS&ME ................................................................................................................ 13 1.1.2 Objective of this report .................................................................................................................. 13
2 USE CASES AND PRIORITY SCENARIOS – SETTING THE SCENE ................................................................ 14
3 ADAS&ME WORK PROCESS ................................................................................................................... 17
3.1 SYSTEM ARCHITECTURE AND SPECIFICATIONS ..................................................................................................... 17 3.2 ENVIRONMENTAL SENSING............................................................................................................................. 19 3.3 DRIVER/RIDER STATE MONITORING.................................................................................................................. 21
3.3.1 Physical Fatigue ............................................................................................................................. 22 3.3.2 Rest ................................................................................................................................................ 22 3.3.3 Sleepiness ...................................................................................................................................... 23 3.3.4 Stress ............................................................................................................................................. 23 3.3.5 Distraction ..................................................................................................................................... 23 3.3.6 Emotion ......................................................................................................................................... 23
3.4 HMI ACTIONS AND TRANSITIONS.................................................................................................................... 25 3.4.1 Iterative HMI development and testing ........................................................................................ 25 3.4.2 HMI Framework............................................................................................................................. 26 3.4.3 HMI Elements and Modalities ....................................................................................................... 27 3.4.4 Automation ................................................................................................................................... 29 3.4.5 HMI personalisation ...................................................................................................................... 30 3.4.6 Decision and Support System ........................................................................................................ 31 3.4.7 Main Innovations........................................................................................................................... 32 3.4.8 Outlook .......................................................................................................................................... 32
3.5 INTEGRATION IN DEMONSTRATORS ................................................................................................................. 33 3.6 EVALUATIONS ............................................................................................................................................. 35
3.6.1 Evaluation framework ................................................................................................................... 35 3.6.2 Pre-pilot data collections ............................................................................................................... 36 3.6.3 Final evaluations ........................................................................................................................... 37 3.6.4 UC A - Truck ................................................................................................................................... 37 3.6.5 UC B – Electrical passenger car ..................................................................................................... 38 3.6.6 UC C/D - Conventional passenger car ............................................................................................ 39 3.6.7 UC E/F - Motorbike ........................................................................................................................ 40 3.6.8 UC G – Automated docking at bus stop ......................................................................................... 41 3.6.9 Demanding issues during the final evaluation .............................................................................. 42 3.6.10 Achievements ........................................................................................................................... 42 3.6.11 Recommendations .................................................................................................................... 43
4 USE CASE APPROACH AND ACHIEVEMENTS .......................................................................................... 44
4.1 USE CASE A – TRUCK ................................................................................................................................... 44 4.1.1 Aim ................................................................................................................................................ 44 4.1.2 Approach ....................................................................................................................................... 44 4.1.3 Achievements ................................................................................................................................ 46 4.1.4 Innovations .................................................................................................................................... 48 4.1.5 Limitations ..................................................................................................................................... 48
4.2 USE CASE B - ELECTRICAL VEHICLE .................................................................................................................. 49 4.2.1 Aim ................................................................................................................................................ 49 4.2.2 Approach ....................................................................................................................................... 49 4.2.3 Achievements ................................................................................................................................ 50
4.3 USE CASE C/D - CONVENTIONAL PASSENGER CAR .............................................................................................. 51 4.3.1 Aim ................................................................................................................................................ 51 4.3.2 Approach ....................................................................................................................................... 51 4.3.3 Achievements ................................................................................................................................ 52 4.3.4 Innovations .................................................................................................................................... 53
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 4 of 68 Version 2.1
4.3.5 Limitations ..................................................................................................................................... 53 4.4 USE CASE E/F - MOTORBIKE AND PROTECTIVE GEAR........................................................................................... 53
4.4.1 Aim ................................................................................................................................................ 54 4.4.2 Approach ....................................................................................................................................... 54 4.4.3 Achievements ................................................................................................................................ 54 4.4.4 Limitations ..................................................................................................................................... 55
4.5 USE CASE G - CITY BUS ................................................................................................................................. 55 4.5.1 Aim ................................................................................................................................................ 55 4.5.2 Approach ....................................................................................................................................... 56 4.5.3 Achievements ................................................................................................................................ 56 4.5.4 Innovations .................................................................................................................................... 57 4.5.5 Limitations ..................................................................................................................................... 57
5 ASSESSMENT OF THE IMPACT ............................................................................................................... 58
5.1 SAFETY, MOBILITY AND ENVIRONMENT ............................................................................................................. 58 5.2 ECONOMIC AND SOCIAL IMPACT ...................................................................................................................... 59 5.3 LEGAL AND REGULATORY IMPACT ASSESSMENT .................................................................................................. 60
6 EXPLOITATION OF THE RESULTS ............................................................................................................ 61
7 DISSEMINATION .................................................................................................................................... 63
8 CONCLUSIONS ....................................................................................................................................... 66
9 CONTACTS ............................................................................................................................................. 67
9.1 COORDINATION TEAM................................................................................................................................... 67 9.2 DISSEMINATION MANAGER ............................................................................................................................ 67 9.3 WP LEADERS .............................................................................................................................................. 67 9.4 USE CASE LEADERS ....................................................................................................................................... 67
10 REFERENCES .......................................................................................................................................... 68
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 5 of 68 Version 2.1
Index of Figures FIGURE 1 THE DECISION AND STRATEGY LOOP FOR ADAS&ME .......................................................................................... 11 FIGURE 2 ADAS&ME PROJECT OUTLINE. ....................................................................................................................... 12 FIGURE 3 OVERALL METHODOLOGY FOR DEFINITION AND SELECTION OF FINAL USE CASES OF ADAS&ME. .................................. 14 FIGURE 4 ADAS&ME WORKSHOP FOCUSING ON USE CASES AND SCENARIOS PRIORITISATION. ................................................. 15 FIGURE 5 SELECTED USE CASES...................................................................................................................................... 16 FIGURE 6 DRIVER/RIDER MODEL. .................................................................................................................................. 16 FIGURE 7 FRAMEWORK FOR ADAS&ME SYSTEM SPECIFICATIONS. ..................................................................................... 17 FIGURE 8 ADAS&ME GENERAL SYSTEM ARCHITECTURE. .................................................................................................. 18 FIGURE 9 DATA FLOW BETWEEN MODULES INCLUDED FOR ENVIRONMENTAL SENSING.............................................................. 20 FIGURE 10 VARIOUS LAYERS OF LOCAL DYNAMIC MAP. .................................................................................................... 21 FIGURE 11 OVERVIEW OF ALL DRIVER STATES AND SENSORS APPLIED FOR EACH USE CASE. ...................................................... 22 FIGURE 12 PERFORMANCE RESULTS OF DRIVER STATES ALGORITHMS. ................................................................................... 24 FIGURE 13 PHYSIOLOGICAL SENSORS DEVELOPED IN ADAS&ME TO COLLECT HEART-RELATED DRIVER DATA UNOBTRUSIVELY IN THE
CAR. LEFT: SENSOR SEAT, RIGHT: RADAR STEERING WHEEL. ...................................................................................... 24 FIGURE 14 ITERATIVE HMI TESTING OVERVIEW FOR ALL USE CASES. ..................................................................................... 25 FIGURE 15 IMPRESSIONS FROM THE ITERATIVE AND USER-CENTRED TESTING OF ADAS&ME HMI. ........................................... 26 FIGURE 16 ADAS&ME “STATECONS” REPRESENTING THE DRIVER STATES. ........................................................................... 26 FIGURE 17 HMI ELEMENTS ALLOCATED TO THE DEMONSTRATOR VEHICLES AND DRIVER STATES. ................................................ 27 FIGURE 18 UC A – HMI ELEMENTS FOR THE TRUCK. ........................................................................................................ 28 FIGURE 19 UC B – HMI ELEMENTS FOR THE ELECTRIC CAR. ............................................................................................... 28 FIGURE 20 UC C/D HMI ELEMENTS FOR THE COMBUSTION ENGINE CAR. ............................................................................. 28 FIGURE 21 UC E/F - HMI ELEMENTS FOR THE PTW. ....................................................................................................... 29 FIGURE 22 UC G - HMI ELEMENTS FOR THE CITY BUS. ...................................................................................................... 29 FIGURE 23 HIGH-LEVEL ARCHITECTURE OF THE PERSONALIZED HMI. .................................................................................... 31 FIGURE 24 DECISION SUPPORT SYSTEM IN THE ARCHITECTURE OF THE COMBUSTION ENGINE CAR. ............................................. 31 FIGURE 25 DECISION SUPPORT SYSTEM IMPLEMENTED INTO THE COMBUSTION ENGINE CAR. .................................................... 32 FIGURE 26: PHASES OF ADAS&ME TECHNICAL VERIFICATION ........................................................................................... 33 FIGURE 27: SCREENSHOTS FROM INTEGRATION PLUGFESTS. ............................................................................................... 34 FIGURE 28: ADAS&ME UC C&D AND UC E&F HMI DEMONSTRATIONS DURING THE 2ND EUCAD CONFERENCE ...................... 34 FIGURE 29: ADAS&ME DEMONSTRATORS DURING THE PROJECT FINAL EVENT. ................................................................... 35 FIGURE 30 OVERVIEW OF 3-PHASE TESTING PROCESS. ....................................................................................................... 36 FIGURE 31 INTERIOR OF SCANIA TEST TRUCK (LEFT) STATIONARY DEMONSTRATION SYSTEM (RIGHT). .......................................... 38 FIGURE 32 ADAS TEST TRACK AT IDIADA ...................................................................................................................... 38 FIGURE 33 OPEN ROAD TEST ROUTE USED FOR UC B. ....................................................................................................... 39 FIGURE 34 PARTICIPANT DRIVING UC B DEMONSTRATOR VEHICLE ON THE TEST ROUTE WITH HMI ACTIVE. ................................. 39 FIGURE 35 USE CASE C DEMONSTRATOR VEHICLE WITH INTEGRATED ADAS&ME COMPONENTS. ............................................. 40 FIGURE 36 GENERAL ROAD TEST ROUTE USED FOR UC E PHYSICAL FATIGUE INDUCEMENT. ....................................................... 40 FIGURE 37 PARTICIPANT WITH DUCATI MULTISTRADA WITH INTEGRATED ADAS&ME COMPONENTS. ....................................... 41 FIGURE 38 VTI BUS SIMULATOR USED FOR THE USE CASE G EVALUATION. ............................................................................ 41 FIGURE 39 PARTICIPANT DRIVING THE VTI BUS SIMULATOR IN USE CASE G. .......................................................................... 42 FIGURE 40 SUMMARY OF UC A ARCHITECTURE................................................................................................................ 45 FIGURE 41 PHOTO OF STAGE I (LEFT) AND STAGE II (RIGHT) DATA COLLECTION. ..................................................................... 45 FIGURE 42 HMI STRATEGY OVERVIEW TABLE. ................................................................................................................. 46 FIGURE 43 GRAPHICAL REPRESENTATION OF HMI ELEMENTS (LEFT PANEL) AND PHOTO FROM SIMULATOR TESTING (RIGHT). .......... 46 FIGURE 44 THE UC C/D CAR. ...................................................................................................................................... 52 FIGURE 45 RIDER PROTECTIVE GEAR WITH INTEGRATED SENSORS. ........................................................................................ 54 FIGURE 46 USE CASE G SCENARIO WITH HMI FOR AN AUTOMATED DOCKING FUNCTIONALITY INTEGRATED ................................. 56 FIGURE 47 IMPACT ASSESSMENT AND ITS RELATIONS TO OTHER PARTS OF THE WORK IN ADAS&ME ......................................... 58 FIGURE 48 SAFETY IMPACT ASSESSMENT PROCEDURE ........................................................................................................ 59 FIGURE 49 ADAS&ME PROJECT YOUTUBE CHANNEL. ...................................................................................................... 63 FIGURE 50 PARTICIPANTS AT THE ADAS&ME FINAL EVENT DECEMBER 2019. ..................................................................... 64
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 6 of 68 Version 2.1
Index of Tables TABLE 1. THE 11 OBJECTIVES OF ADAS&ME.................................................................................................................. 13 TABLE 2 DIFFERENT TYPES OF RISKS IDENTIFIED IN ADAS&ME ........................................................................................... 19 TABLE 3 AUTOMATED FUNCTIONS IN THE HMI STRATEGY PER USE CASE................................................................................ 30 TABLE 4 EXPLOITABLE RESULTS OF ADAS&ME ............................................................................................................... 61
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 7 of 68 Version 2.1
Glossary
AD AUTOMATED/ AUTONOMOUS DRIVING
AV AUTOMATED/ AUTONOMOUS VEHICLE
ADAS ADVANCED DRIVER ASSISTANCE SYSTEM
API APPLICATION PROGRAM INTERFACE
ARAS ADVANCED RIDER ASSISTANCE SYSTEM
CPM COOPERATIVE PERCEPTION MESSAGE
DSS DECISION SUPPORT SYSTEM
DOA DESCRIPTION OF ACTION
EEG ELECTROENCEPHALOGRAM
FQMR FINANCIAL QUARTERLY MONITORING REPORT
HD HIGH DEFINITION
CAD CONNECTED & AUTOMATED DRIVING
DIM DRIVER IDENTIFICATION MODULE
ECG ELECTROCARDIOGRAM
FMEA FAILURE MODE AND EFFECTS ANALYSIS
GDPR GENERAL DATA PROTECTION REGULATION
GSA GENERAL SYSTEM ARCHITECTURE
GSR GALVANIC SKIN RESPONSE
GUI GRAPHICAL USER INTERFACE
HD MAP HIGH DEFINITION MAP
HMI HUMAN MACHINE INTERACTION
HRV HEART RATE VARIABILITY
IPR INTELLECTUAL PROPERTY RIGHTS
ITS INTELLIGENT TRANSPORTATION SYSTEM
KPI KEY PERFORMANCE INDICATOR
KSS KAROLINSKA SLEEPINESS SCALE
MQTT MQ TELEMETRY TRANSPORT (A MACHINE-TO-MACHINE (M2M)/"INTERNET OF
THINGS" CONNECTIVITY PROTOCOL)
NDA NON-DISCLOSURE AGREEMENT
PTW POWERED TWO-WHEELER
QMR QUARTERLY MONITORING REPORT
UC USE CASE
VR VIRTUAL REALITY
WP WORK PACKAGE
LED LIGHT-EMITTING DIODE
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 8 of 68 Version 2.1
MCA MULTI-CRITERIA ANALYSIS
PHP HYPERTEXT PRE-PROCESSOR
PPE PERSONAL PROTECTIVE EQUIPMENT
PS PERSONALISATION SYSTEM
PPS PRE-PILOT STUDIES
REST REPRESENTATIONAL STATE TRANSFER
SOA STATE OF THE ART
SQL STRUCTURED QUERY LANGUAGE
XML EXTENSIBLE MARK-UP LANGUAGE
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 9 of 68 Version 2.1
Executive Summary
ADAS&ME (“Adaptive ADAS to support incapacitated drivers & Mitigate Effectively risks
through tailor made HMI under automation”) aimed to develop Advanced Driver Assistance
Systems that incorporate driver/rider state, situational/environmental context, and adaptive
interaction to automatically transfer control between vehicle and driver/rider and thus ensure
safer and more efficient road usage for all vehicle types (conventional and electric car, truck,
bus, motorcycle). To achieve this, a holistic approach has been applied which considers
automated driving along with information on driver/rider state and the environment.
This report is the Final report of ADAS&ME with the aim to provide the reader with an
overview of the ADAS&ME project scope, the outline of the work, how it was performed, and
an overview of the results achieved.
Chapter 1 describe the overview of the scope of ADAS&ME and its objectives, chapter 2 focus
on the use cases and the scenarios that based on indicative use cases were updated and for which
the prioritisation of scenarios was made. The use cases were then matched with demonstrators
in which all systems develop had to be integrated. The demonstrators were also used for the
final evaluation. Chapter 3 cover the work process of ADAS&ME including the steps system
architecture and specification, environmental sensing, driver/rider state assessment, HMI action
and transition, integration and final evaluations. Chapter 4 then gives an overview of the
approach and achievements for each use case. Chapter 5 describe the impact assessment.
Chapter 6 describe the exploitation and chapter 7 gives a brief overview of the disseminations
of ADAS&ME. Chapter 8 is about the conclusions in relation to the objectives of ADAS&ME
and finally chapter 9 provide a list of contact persons per work package and use cases.
ADAS&ME included 11 objectives with the following overall achievements: ▪ The development of robust detection/prediction algorithms for driver/rider state enabling
personalisation of individual driver’s physiology and driving behaviour. This has been
achieved with good accuracy for mostly all driver state detections.
▪ The development of multi-modal, user oriented and adaptive information, warning, actuation
and handover strategies, based on current and predicted driver/rider state, criticality of
scenario and its environmental context. This has been achieved for all seven use cases.
▪ Integration of the developed algorithms, sensing technologies, supportive technologies
(automation, V2X) and HMI algorithms/components into driver/rider state monitoring
systems. This has been achieved in all five demonstrators.
▪ The development of personalised driver/rider behaviour profiles, considering inter-individual
differences. This has been achieved in the majority of the use cases.
▪ A design of HMI concepts, prototypes and guidelines for automated functions that take
driver/rider state into account, for possible implementation in future driver/rider state-adapted
automated systems. This has been achieved for all seven use cases.
▪ Instrumentation of evaluation/demonstration tools (simulators and vehicles) for evaluation of
the developed systems in different environments, as specified by the ADAS&ME use cases
and its adaptive architecture. This has been achieved for all seven use cases expect for those
that due to high safety risk with driver state inducement had to be simulated to evaluate the
HMI components.
▪ Adaptation of existing EuroNCAP test protocols from non-automated to automated driving
modes. The experience in the project has been used as input for future test protocols.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 10 of 68 Version 2.1
▪ Perform targeted tests for the selection of HMI elements that optimally support each Use Case.
This has been done for all use cases as a part of an agile development including iterative
testing.
▪ Ethical and legal considerations regarding the experiments carried out on driver/rider state
monitoring, the evaluation of the developed systems, as well as when and how automated
systems should interact with the driver/rider. This has been achieved and all test with humans
have been ethical approved and in line with the legalisation.
▪ An evaluation of the developed systems and use cases with a wide pool of drivers/riders under
simulated, controlled and real road conditions and for different driver/rider states and
automation use cases/levels. This has been achieved with almost 200 drivers/riders involved in
data collection for algorithm and evaluations.
▪ A holistic impact assessment of automation opportunities to enhance safety by supporting the
impaired driver/rider, as well as of handover transitions optimisation, taking the driver/ rider
estimated state into account. This has been achieved for all seven use cases.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 11 of 68 Version 2.1
1 The scope of ADAS&ME
ADAS&ME (“Adaptive ADAS to support incapacitated drivers & Mitigate Effectively risks
through tailor made HMI under automation”) aimed to develop Advanced Driver Assistance
Systems that incorporate driver/rider state, situational/environmental context, and adaptive
interaction to automatically transfer control between vehicle and driver/rider and thus ensure
safer and more efficient road usage for all vehicle types (conventional and electric car, truck,
bus, motorcycle). To achieve this, a holistic approach has been applied which considers
automated driving along with information on driver/rider state and the environment.
The holistic approach of ADAS&ME considers automated driving/riding along with
information on driver/rider state, to develop optimized HMI and support strategies, where
automated and partly automated driving/riding is seen as both an influencing factor and a tool
to affect driver/rider state, see Figure 1.
In the ADAS&ME decision and strategy loop the driver state is influenced by different factors,
including individual factors (e.g. sleepiness, inattention, stress, anxiety, physical fatigue etc.),
external factors (e.g. weather conditions, traffic density/interactions with others) and the
driver/rider vehicle interaction mode (automated, semi-automated or manual). Sometimes
driving/riding performance, like speed, lane keeping, or headway will be influenced by the
driver/rider state, as well as by the driver/rider vehicle interaction mode and by external factors.
Based on the output from driver/rider state monitoring and current driving performance, a
decision about the need of mitigation or countermeasures is made and used as input to the
strategy for intervention or driver vehicle interaction change. The strategy uses information,
warning, actuation and handover strategies, depending on the need, as well as different HMI
modalities (haptic/tactile, auditory and visual).
Figure 1 The decision and strategy loop for ADAS&ME
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 12 of 68 Version 2.1
The work in ADAS&ME is built around 10 Work Packages (WP), see Figure 2. In addition,
use case teams have been working in parallel. The starting point for all work is a generic
approach that humans with their pros and cons benefit and suffer from the same things
regardless of what type of vehicle they are driving. This means that generic work on system
architecture, specifications and verifications are used, but also that driver/rider state detection
and HMI strategies are based on the same principles, but of course adjusted per use case and its
demonstrator.
Figure 2 ADAS&ME project outline.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 13 of 68 Version 2.1
1.1 Objectives
1.1.1 Objectives of ADAS&ME
ADAS&ME encompasses 11 key objectives (all achieved), see Table 1.
Table 1. The 11 objectives of ADAS&ME
Objectives
Development of robust detection/prediction algorithms for driver/rider state monitoring
of fatigue/drowsiness, stress, inattention/distraction and impairing emotions, employing
14 existing and novel sensing and speech technologies, thereby enabling ADAS&ME to
be personalised for individual driver’s physiology and driving behaviour.
Development of multi-modal, user oriented and adaptive information, warning, actuation
and handover strategies, based on current and predicted driver/rider state, criticality of
scenario and its environmental context.
Integration of the developed algorithms, sensing technologies, supportive technologies
(automation, V2X) and HMI algorithms/components into driver/rider state monitoring
systems.
Development of personalised driver/rider behaviour profiles, considering inter-
individual differences.
Design of HMI concepts, prototypes and guidelines for automated functions that take
driver/rider state into account, for possible implementation in future driver/rider state-
adapted automated systems.
Instrumentation of evaluation/demonstration tools (simulators and vehicles) for
evaluation of the developed systems in different environments, as specified by the
ADAS&ME use cases and its adaptive architecture.
Adaptation of existing EuroNCAP test protocols from non-automated to automated
driving modes.
Performance of targeted tests for the selection of HMI elements that optimally support
each Use Case.
Ethical and legal considerations regarding the experiments carried out on driver/rider
state monitoring, the evaluation of the developed systems, as well as when and how
automated systems should interact with the driver/rider.
Evaluation of the developed systems and use cases with a wide pool of drivers/riders
under simulated, controlled and real road conditions and for different driver/rider states
and automation use cases/levels.
Holistic impact assessment of automation opportunities to enhance safety by supporting
the impaired driver/rider, as well as of handover transitions optimisation, taking the
driver/ rider estimated state into account.
1.1.2 Objective of this report
This report is the Final report of ADAS&ME with the aim to provide the reader with an
overview of the ADAS&ME project scope, the outline of the work, how it was performed, and
an overview of the results achieved.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 14 of 68 Version 2.1
2 Use cases and priority scenarios – setting the scene
ADAS&ME is built around use cases and scenarios. Already from the beginning of the project
seven indicative use cases was defined. At the beginning of the project, those were further
analysed, and a set of relevant scenarios were developed for each of them. Each use case was
then matched to a specific demonstrator in which all components dealing with environmental
sensing, detection of driver/rider state and HMI develop were integrated. The demonstrators
were also used for the final evaluation phase.
The selection of use cases and the prioritisation of scenarios aimed to set the scene. This
required a work to identify the most important issues and knowledge gaps related to driver state
and automated driving, to understand the users’ needs (driver/riders), see Figure 3. The work
started with an SoA and benchmarking activity (Touliou, Maglavera, & Britsas, 2017).
Templates were prepared and shared with partners to gather information about available
solutions for driver/rider monitoring. A template was also prepared and filled out to align
actions to specific priorities.
Figure 3 Overall methodology for definition and selection of final use cases of
ADAS&ME.
End users’ perspectives and wishes were gathered through an open web-survey, and an analysis
was done based on more than 1000 EU respondents. The survey was built around short stories
for each use case that together with illustrations were used in the web survey to visualize the
concept. The web survey used the SoSci Survey tool (https://www.soscisurvey.de/en/about
retrieved: 2020-02-21). A wide range of stakeholders, identified with the help of input from all
partners, were invited. To complement the survey five focus groups were organised. The aim
was to get a deeper understanding of the users’ views and wishes. In addition, a workshop with
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 15 of 68 Version 2.1
participation of more than 25 stakeholder representatives was held with the aim to provide input
to the final selection of the use cases and prioritisation of scenarios per use case, see Figure 4.
Figure 4 ADAS&ME workshop focusing on Use Cases and scenarios prioritisation.
The prioritisations from the different quantitative methodologies were consolidated with help
of a Multi-Criteria Analysis (MCA) based on all stakeholders input and a final prioritisation of
scenarios for each use case was achieved. This work also included an analysis of the potential
impact of the selected use cases and scenarios. In total seven use cases were selected with two
priorities per identified use case (Dukic Willstrand et al., 2017). In the end, seven ADAS&ME
use cases were selected for implementation together with two prioritised scenarios, see Figure
5.
The prioritised scenarios for each use case were:
Use case A: Attentive long-haul trucking (truck) ▪ 1st Priority: Safe Stop
▪ 2nd Priority: Traffic Jam negotiation
Use case B: Electric Vehicle (EV) range anxiety (e-car) ▪ 1st Priority: EV range problem appears due to traffic jam
▪ 2nd Priority: Driver trusts the system and follows indications to power charging station
Use case C: Drive state based smooth and safe automation transitions (conventional car) ▪ 1st Priority: Unsuccessful handover due to non-reacting driver and safe stop
▪ 2nd Priority: Controlled (automation initiated) handover transitions taking the driver state into
account for the interaction design into
Use case D: Non-reacting driver emergency manoeuvre (conventional car) ▪ 1st Priority: Unsuccessful handover and takeover transitions
▪ 2nd Priority: Controlled (automation initiated) takeover transitions based on driver state
Use case E: Long range attentive touring with motorbike (motorbike) ▪ 1st Priority: Assistance during long range touring in case of tiredness
▪ 2nd Priority: Assistance during long range touring in case of inattention
Use case F: Rider Faint (motorbike) ▪ 1st Priority: Activation of active systems if the rider is fainting
▪ 2nd Priority: Activation of active systems if the rider is going to faint and is ignoring
assistance
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 16 of 68 Version 2.1
Use case G: Passenger pick-up/drop-off automation for buses (city bus) ▪ 1st Priority: System initiated takeover
▪ 2nd Priority: Driver initiated takeover
Figure 5 Selected use cases.
Note: from left to right: UCA Long haul trucking, UCB electrical vehicle range anxiety, UCC
and UCD Conventional car: “driver state based safe and smooth automation transition” and
“none reacting driver emergency manoeuvre”, UCE and UCF “long range attentive touring with
motorbike” and “rider faint”, UCG Passenger pick up/drop off automation for buses.
A theoretical model for driver/rider state was also defined (Dukic Willstrand et al., 2017).
General definitions of the driver/rider states, investigated in ADAS&ME, were elaborated in
collaboration with the developers of the driver/rider state detection algorithms, Figure 6.
Suitable reference/ground truth measures and performance evaluation methods for the
individual states were investigated and research on ground truth generation methods and
suitable experiments for “stress/workload”, “inattention/distraction” and “physiological
impairment” were conducted.
Figure 6 Driver/rider model.
The work with “setting the scene” was performed mainly in Work Package 1 of ADAS&ME.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 17 of 68 Version 2.1
3 ADAS&ME work process
3.1 System architecture and specifications
Work package 2 focused on ADAS&ME System architecture and specifications. The goal of
this work was to design a flexible architecture to facilitate smooth integration of various
software and hardware components designed during the project across the identified use cases.
This work consisted of three major activities:
1. System Architecture
2. Technical specifications
3. Risk Analysis
Figure 7 shows the framework followed for achieving system specifications. Once the scope of
the use cases was finalized, the work in WP2 started by collecting the requirements for the use
cases.
System Design
ADAS&ME Project Proposal Idea
Survey
State of the Art List of Users User wishes Stakeholder wishes
Needs and Limitations
User needs Stakeholder needs Planned Innovation Limitations Use Cases
System Requirements
Tasks Communication Timing Priorities
Functional Blocks
ADAS&ME System Specifications
Input Data Output Data
Error Handling
Figure 7 Framework for ADAS&ME System Specifications.
Based on the collected requirements, the ADAS&ME General System Architecture was
designed, see Figure 8. This led to the visualization of key elements (e.g. subsystems, modules
and components) that are required for a successful execution of the use cases in the final
demonstrators. These general key elements were modified to suit the needs of each
demonstrator. System architecture was specified for each use case at different levels e.g.
functional architecture, communication architecture and the physical architecture. In order to
facilitate a flexible communication across various heterogeneous elements, ADAS&ME GSA
presented an innovative way by unifying the message data used for communication across
various components.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 18 of 68 Version 2.1
The technical specification of each key element was identified using a template. The template
described the key elements in a tabular form and consisted of 4 sections, namely:
1. Introduction – containing the information about the manufacturer, input and output
details.
2. Performance section – containing information about the key performance aspects of the
elements e.g. accuracy of measurement, frequency etc. of a sensor.
3. Physical aspects section – containing the information about the physical weight and
dimension of the elements.
4. Environmental aspects section – containing the information about the tolerable
environmental conditions e.g. temperature for the respective element to function
correctly.
Environmental Situation Awareness Subsystem (ESAS)
Algorithms for Environmental Situation Determination
ADAS&ME Core (ADAS&ME C)
Personalization Module (PM)Decision Support Module (DSM)
HMI Controller Module (HMI CM)
Vehicle Automation Subsystem (VAS)
Vehicle Automation Module (VAM)
Soft Sensor data
Driver monitoring Data
Environmental/Vehicle Data
Digital Infrastructure
DataHMI Data
Interface Module (IM)
Sensor Subsystem (SS)
Driver State Monitoring Subsystem (DSMS)
Algorithms for Driver State Determination
Figure 8 ADAS&ME General System Architecture.
A risk assessment was carried out using Failure Mode and Effects Analysis
(FMEA)methodology to identify the design flaws and any issues that could occur during the
integration and execution of the use cases. Risks were identified in 4 categories, i.e. technical,
behavioural, legal and organizational. Table 2 quantitatively describes the various risks
identified during the project.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 19 of 68 Version 2.1
Table 2 Different types of risks identified in ADAS&ME
3.2 Environmental sensing
Within the scope of ADAS&ME, Advanced Driver Assistance Systems (ADAS) were
developed in order to ensure safer and more efficient road usage. These ADAS incorporate
driver/rider state, situational/environmental context, and adaptive interaction to automatically
transfer control between vehicle and driver/rider. The goal with the situational/environmental
was to derive an indicator about the stressfulness or dangerousness of the surrounding. The
indicator is supposed to serve a following decision-making module in deciding whether the
current environmental situation is advantageous for manual or autonomous driving and whether
a transfer of control would be beneficial.
Figure 9 shows all modules that have been developed. The Sensor Data Fusion Module
processes sensor data to derive the own vehicle state. The Communication Module consists of
three sub-modules: The Digital Infrastructure looks further along the road ahead to understand
what to expect. V2X Data Sharing provides means to communicate with the close vicinity via
V2X communication. The information received via V2X is fused by the V2X Data Integration.
The Environmental State Determination Module combines in a first step the own vehicle state
and environmental context information in order to form a Local Dynamic Map (LDM). Step 2
estimates the behaviour of objects present in the LDM over a specified time and classifies the
current environment situation as Normal, Critical or Dangerous.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 20 of 68 Version 2.1
Figure 9 Data flow between modules included for environmental sensing.
The Sensor Data Fusion Module determines the vehicle’s current location in a reliable and
probabilistic way. The innovation compared to state-of-the-art is that the confidence of the
location estimation itself is estimated. For next-generation automated driving functionality it is
important to not only know the position itself, but also the confidence.
The Digital Infrastructure determines the driver task intensity based on static and transient
dynamic data. More precisely is the driving task intensity derived from traffic flow data,
location-based weather data and High Definition (HD) map data such as number of lanes, lane
transitions and lane geometry (e.g. curvature). The driving task intensity quantifies the
stressfulness and dangerousness of the road ahead ignoring the highly dynamic data.
The V2X Data Sharing covers all (vehicle external) communication aspects. It enables direct
vehicle to vehicle and vehicle to infrastructure communication according to ETSI standards
(https://www.etsi.org/ Retrieved:2020-02-21). As part of the ADAS&ME activities, new
messages for collaborative perception and cooperative manoeuvres were developed and
discussed in ETSI.
The V2X Data Integration fuses information that is received by the V2X Data Sharing and the
measurements from the on-board sensors. This includes a new proposal for the temporal and
spatial alignment, innovative CPM dissemination strategies and concepts for more robust
perception.
The task of the Environmental State Determination Module is to assess the current
environmental situation around the own vehicle. Data describing the situational/environmental
context is provided by the aforementioned modules. Distinct sources such as maps, landmarks,
weather information, and vehicles or infrastructure that are equipped with exteroceptive and
proprioceptive sensors have been considered. In a first step the data is gathered by the Local
Dynamic Map (LDM), which stores the information depending on the persistency of the
information. As shown in Figure 10 the LDM had four layers with different types of data.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 21 of 68 Version 2.1
Figure 10 Various Layers of Local Dynamic Map.
The LDM enables awareness of the current situation around the automated vehicle From this
point, and in order to assess the situation, a Decision Network (DN) was designed. It considers
estimated values from the sensors along with their confidence values and computes the utility
value for each decision alternative. The principle of maximum expected utility is applied i.e.,
the decision alternatives of the environment with highest utility will be chosen as output.
The work on “environmental sensing” were performed mainly in Work Package 3 of
ADAS&ME.
3.3 Driver/rider state monitoring
The objective of the driver/rider state monitoring was to develop and demonstrate algorithms
and sensors for driver/rider state detection. Within the context of ADAS&ME these solutions
should be able to detect and measure undesirable or unusual driver conditions to enhance safety
of drivers/riders. The drivers were supported during handover transitions thanks to the
developed adaptive HMI systems considering the actual driver state.
After receiving the input from the stakeholder and end user surveys, which resulted in an update
of the use case scenarios, an updated list of driver states per Use Case was generated. Within
this work, indicators for each of the specified driver states were identified. Furthermore, a
thoroughly defined theoretical background for each driver state was established which is
documented in deliverable 4.2 (Hennes & Mathissen, 2020) (confidential). The work was
supported by the driver/rider state model (Dukic Willstrand et al., 2017). Additional
contributions were coming from expert background knowledge, literature review and additional
acquisition of data from testing both in lab and in the field. The driver state indicators were then
used for the development and evaluation of the actual driver state detection algorithms. To
provide input data to the driver state algorithms specific sensors were required. A state-of-the-
art for sensors were developed and an identification of high-quality data for the driver state
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 22 of 68 Version 2.1
algorithms were made. During the project, different sensors were tested and either rejected or
verified to be applicable in the specific use cases which resulted in a final set of sensors for
each use case. The sensors and algorithms were later integrated into the demonstrators and
evaluated with all other ADAS&ME components as a complete system.
An overview of all selected driver/rider state sensors and algorithms is given in Figure 11.
Figure 11 Overview of all Driver states and Sensors applied for each Use Case.
A total of 7 different sensor systems provide data to algorithms detecting 6 different driver
states: Distraction, Emotions, Rest, Physical Fatigue, Sleepiness and Stress. These were
integrated in the five demonstrators covering all seven use cases. Main innovations and
limitations for each of the driver/rider state algorithms are several.
3.3.1 Physical Fatigue
For the first time a model for predicting thermal fatigue of riders was developed. This was
enabled by a unique set of data collected during the simulation tests on rider thermal stress.
Since there is almost no background research on rider state detection a lot of innovative work
was done within the rider use case (UC E/F) in ADAS&ME.
3.3.2 Rest
For the first time a driver state algorithm estimating drivers’ rest was developed. A main
challenge was the missing ground truth of resting which posed a challenge for the developers.
A sophisticated three-process model of alertness was used to estimate how much driving time
has been gained by resting.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 23 of 68 Version 2.1
3.3.3 Sleepiness
This driver state has been very well investigated in the last decades. Within the project, a new
representation of input features was introduced and specifically designed to improve
performance when using neural networks.
3.3.4 Stress
Three very different stress inducement strategies were hypothesized and finally tested to select
the best one for final evaluation. It was shown that personalization of the algorithm shows
improvement over the generalized algorithmic approach (ca. +10% F1-Score).
3.3.5 Distraction
The same algorithmic approach was used for different UCs which was further enhanced by
tailoring to the specific features of the demonstrator vehicles. For the truck use case (UC A) the
importance of mirrors was acknowledged while for the bus use case (UC G) a novel situation-
dependent multi-buffer approach was developed. For the rider use case (UC E/F), the first ever
distraction algorithm for riders was developed.
3.3.6 Emotion
Three different emotion algorithms were developed (audio, video and physiological based). The
combination of different modalities allowed continuous tracking of the emotional state. The
algorithms allowed classification of low-expressive emotions of high naturalness.
Generally, different driver/rider states share certain commonalities, as:
• It was often difficult to collect relevant data due to safety or ethical reasons.
• In many cases there was a high inter-individual variability which would require large
data sets to be addressed properly.
• For most driver states, the problem of finding a proper ground truth which is less
intrusive than questionnaires during driving persists.
An overview of the performance metrics of each algorithm is shown in Figure 11.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 24 of 68 Version 2.1
Figure 12 Performance results of driver states algorithms.
Note: Data points refer to mean values and error bars to standard deviation while coloured
background indicates the range. The number of prediction levels (classes) per algorithm is
indicated in parenthesis.
There were two different classes of sensors implemented in the ADAS&ME project.
First, off-the-shelf sensors were chosen based on initial requirements and evaluated for their
actual performance by lab testing. Secondly, new contactless sensors for heart-related
measurements were developed and designed to be directly integrated into the accessories of
the car, either in the driver’s seat or in the steering wheel of the vehicle as shown in Figure 13.
Figure 13 Physiological sensors developed in ADAS&ME to collect heart-related driver
data unobtrusively in the car. Left: Sensor seat, Right: Radar steering wheel.
Furthermore, a novel wearable sensor system designed for riders was developed and directly
integrated into rider’s equipment. While the contactless sensors were working reasonably well
under laboratory conditions, their performance in a moving vehicle was significantly degraded.
Overall, unobtrusive sensors for capturing physiological data remain a major challenge for
driver state detection and need further improvement. Within ADAS&ME, reference sensors
were used as a fall-back solution for UCs that planned to rely on unobtrusive sensors.
In addition to algorithms and sensors, a Personalisation system (PS) was developed consisting
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 25 of 68 Version 2.1
of two distinctive modules. The “Driver Identification Module (DIM)” allows identification of
drivers/riders and the “Personalization module (PM)” gathers, stores and provides relevant
information. While the PM follows the same approach across all UCs the DIM is dependent on
the UC, e.g. the driver/rider may be identified by his/her face (via the SEP camera) or personal
item (rider suit, mobile phone, etc.). The PS was designed as a cross-WP system, e.g. the
personalization of HMI modalities is based on this system. All algorithms, sensors and the
personalization system communicate via MQTT which was defined as a message protocol.
Other work packages adapted this protocol afterwards which makes it the common exchange
platform for messages. Furthermore, all algorithms and sensors follow the same message format
across all platforms.
The work on “Driver/Riders state detection” was performed mainly in work package 4 of
ADAS&ME.
3.4 HMI Actions and Transitions
The aim was to develop HMI elements and strategies for adaptive transitions, a work done in
WP5. HMI elements, adaptive HMI actions and adaptive transitions were developed, evaluated
and demonstrated in all envisioned demonstrators, namely truck, electric car, combustion
engine car, PTW and city bus.
3.4.1 Iterative HMI development and testing
Due to the diversity of the demonstrator vehicles and use cases the approach was to develop
different HMI elements, modules and strategies, following a common HMI framework and
design recommendations. All developments also followed a common user-centred development
process with several development iterations and intermittent user tests. Figure 14 shows the
iterations carried out in the HMI development for each use case and the equipment used in the
laboratories of the responsible partners.
Figure 14 Iterative HMI testing overview for all use cases.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 26 of 68 Version 2.1
Figure 15 shows impressions from the iterative testing of the HMI developed or adapted in
ADAS&ME. In total, feedback and user experiences of 237 potential users of the technology
have been included in the development process.
Figure 15 Impressions from the iterative and user-centred testing of ADAS&ME HMI.
The achievements are different HMI elements, adapted to the project needs or specifically
developed within the project, tailored HMI strategies for each use case and multiple driver
states, the user-centred iterative development, and the final evaluation and demonstration in the
demonstrator vehicles. Common HMI elements have been used where meaningful, especially
related to the driver states and the levels of automation which were common among the use
cases. As a prominent example, “statecons” were developed and used in all use cases in order
to represent the driver states, see Figure 16.
Figure 16 ADAS&ME “Statecons” representing the driver states.
3.4.2 HMI Framework
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 27 of 68 Version 2.1
The different driver states required very different HMI strategies. For example, a moderately
sleepy driver requires interaction and involvement in the driving task, while a very sleepy driver
requires a rest and if possible, Level 4 automation (https://www.sae.org/). A stressed driver
requires little disturbance and a nudging reminder to use automation, while a distracted driver
requires an earlier prompt to take control when using automation. The HMI strategies are
represented in a generic framework possible to use to be adapted to define the strategies for all
target vehicles. The framework enables defining the vehicle–system–user
interactions, considering the driver/rider state, personalization, the environmental context and
all HMI elements (Diederichs et al., 2018).
Besides the different HMI strategies also the different HMI elements can be used for different
driver states. Figure 17 shows the allocation of HMI elements to the use cases and its
demonstrator, considering also the driver/rider states.
Figure 17 HMI elements allocated to the demonstrator vehicles and driver states.
3.4.3 HMI Elements and Modalities
The HMI elements and modalities for each demonstrator are represented in Figure 18 - Figure
22.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 28 of 68 Version 2.1
Figure 18 UC A – HMI elements for the truck.
Figure 19 UC B – HMI elements for the electric car.
Figure 20 UC C/D HMI elements for the combustion engine car.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 29 of 68 Version 2.1
Figure 21 UC E/F - HMI elements for the PTW.
Figure 22 UC G - HMI elements for the city bus.
3.4.4 Automation
In ADAS&ME the automated functions were understood as an important part of the HMI
strategy as they provide important kinaesthetic information to the drivers and passengers. Thus,
the actuation of automation needs to be well explained and seamlessly integrated by the HMI
elements and strategies. Table 3 displays the automated functions that were realized for the final
demonstration.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 30 of 68 Version 2.1
Table 3 Automated functions in the HMI strategy per use case.
Use Case Automated functions in the HMI strategy:
Driver state related automation behaviour
Use Case A
(Demonstrator: Truck)
Safe Stop
Use Case B
(Demonstrator: Electric
vehicle)
Transition to Automation at critical battery level,
Automated driving to next car sharing, Automated
parking
Use Case C
(Demonstrator:
Passenger Vehicle)
Adaptation regarding timing and information strategy in
transitions of control (SAE3-SAE0;
https://www.sae.org/)
Use Case D
(Demonstrator:
Passenger Vehicle)
Timing of Minimum Risk manoeuvre (earlier in critical
driver states)
Use Case E/F
(Demonstrator:
Motorbike)
Adaptation regarding HMI warning with progressive
severity, including torque limitation and capsize control
Use Case G
(Demonstrator: Bus
simulator)
Automated approaching of bus stop, automated stop,
automated departure from bus stop, Safe Stop
3.4.5 HMI personalisation
The adaptation of the HMI also included personalisation of it. For the realization of a
personalized HMI a rule engine with personalisation and adaptation rules was developed. It
enables working in distributed environments with a personalised HMI Controller acting as a
server and HMI modalities and GUI applications acting as clients. Figure 23 shows the high-
level architecture of the personalized HMI.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 31 of 68 Version 2.1
Figure 23 High-level architecture of the personalized HMI.
3.4.6 Decision and Support System
A Decision Support System (DSS) was developed for the decision of which interaction strategy
should be selected in case multiple driver states appear at the same time. Individual software
architectures of the different demonstrator vehicles required individualized solutions, adapted
to existing architectures and OEM specific procedures in the vehicles. Figure 24 show the DSS
implementation in the combustion engine car of UC C/D.
Figure 24 Decision Support System in the architecture of the combustion engine car.
Figure 25 show the implementation of the DSS into the combustion engine car in an example.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 32 of 68 Version 2.1
Figure 25 Decision Support System implemented into the combustion engine car.
3.4.7 Main Innovations
Main innovations of WP5 are related to the development of the first application of new HMI
elements, particularly:
▪ Invention of “Statecons”.
▪ Establishment of LED strips for automation and transition feedback.
▪ Inclusion of automated functions and transitions into HMI strategy.
▪ Personalization framework and adaptation based on driver states.
▪ Consistent HMI strategies for transitions into and out of automation in highly
demanding situations with compromised drivers.
▪ PTW capsize stabilization, rider HMI integrated in gloves and helmet and information
to other road users that impaired rider state is detected.
• First time application of automated functions and driver state adaptive transitions for
Truck, Car, PTW and Bus.
3.4.8 Outlook
The ADAS&ME HMI achievements are highly relevant for further research and series
production vehicles. From a research point of view a driver state adaptive HMI should be
applied also in future projects covering more type use cases, since it will go to series production
timely as from 2022 the EU safety regulation requires it. A main result for future research and
Decision Tree – UC C HandoverH
MI S
trat
egy
Fru
stra
tion
Str
ess
Sle
epin
ess
Dis
trac
tion
1 : Alert
2: No 1 :Yes
3 : Very Drowsy
CoE_S < CoE_D
2 : Drowsy
Sleepiness State
1 : Alert
Sleepiness State
3 : Very Drowsy
Sleepiness State
2 : Drowsy
0:Normal 1:Elevated
3 : Very Drowsy
CoE_S ≥ CoE_D
3 : Very Drowsy
0:Normal 1:Elevated 1:Elevated 0:Normal 0:Normal 1:Elevated 1:Elevated 0:Normal 1:Elevated 0:Normal 0:Normal 1:Elevated
YesNo YesNo No Yes No Yes No Yes YesNo No Yes YesNo
HMI Strategy for distraction HMI Strategy for sleepiness
No Yes No Yes No Yes YesNo NoYes No Yes
HMI Strategy for Stress
* **
* **
HMI Strategy for FrustrationDefault HMI Strategy (No State)
** : CoE_Str<CoE_Frust* : CoE_Str≥CoE_Frust
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 33 of 68 Version 2.1
development is that different driver states and different vehicles require development of
different HMI strategies and HMI elements. For the transition automation to driver more user
tests, legal considerations and ethical debates are needed to discuss if and how drivers shall be
allowed to get control back when their driver state is compromised. For the transition driver to
automation the application of driver state adaptive HMI, supporting the transition from L0 to
L4, is very innovative with a high potential for enhanced safety. It More effort in development
and research as well as in public discussions is required to form an understanding of the need
from the public society.
The relevant work of “HMI action and transitions” was performed mainly in Work package 5
of ADAS&ME.
3.5 Integration in Demonstrators
The architectural framework was used as the starting point to achieve a successful integration
of the developed system into all demonstrators. The framework covers all individual
ADAS&ME components, encompassing the HMI, sensors, algorithms and environmental
monitoring functions. The starting point for the integration was the work done on system
architecture and specifications. The integration work included the task of defining protocols to
be used during the technical verifications of sensors, algorithms, modules and use cases towards
final readiness testing before the launch of the evaluation work performed in the evaluation
work package, see Figure 26. To support the integration framework a platform was developed
on which all subsystem providers (mainly hardware) and those responsible for a demonstrator
added the corresponding information. The tool was used to support the integration and
evaluation phases and to better organise the transfer of components between integration and
evaluation sites as well as to get information on spared components in case of a broken or
problematic component.
Figure 26: Phases of ADAS&ME Technical Verification
Within the framework of the use cases integration a series of ”Plug-fests” (technical integration
meetings) was introduced in order to better monitor but also in the best way facilitate the
cooperation between the use case integration teams, see Figure 27. The final integration at the
evaluation sites took place in order to facilitate the requirements of each use case evaluation.
Phase 1 : Design of ADAS&ME Technical
Verification Plans
Phase 2 : Implementation of ADAS&ME
Technical Verification Plans
Phase 3 : ADAS&ME Technical Verification Results and Analysis
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 34 of 68 Version 2.1
Figure 27: Screenshots from Integration Plugfests.
Furthermore the work included the preparation of Use Case demonstrations in dedicated
external events, such as the UCA demonstration at the 2018 2018 EUCAR Reception and
Conference which took place on 6-7 November in Brussels, using the SCANIA truck VR
simulator and the UC C&D and UC E&F adaptive HMI demonstrations during the 2nd EUCAD
Conference held on April 2-3, 2019 in Brussels, see Figure 28.
Figure 28: ADAS&ME UC C&D and UC E&F HMI demonstrations during the 2nd
EUCAD Conference
In addition the preparation of the demonstrations for the Project Final Event that was held on
December 3rd, 2019 at IDIADA premises was also realised for all the Use Cases of the project,
see Figure 29.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 35 of 68 Version 2.1
Figure 29: ADAS&ME Demonstrators during the Project Final Event.
As an output of this work, a list of recommendations for the future was devised as follows:
• Revisit physical and communication architecture before integration starts.
• Continuous risk assessment at all key phases of the project and at all levels;
technical, operational, behavioural and legal.
• Perform incremental integration; components to systems -> technical verification;
systems to systems -> technical verification and so on to avoid solving an error at a
late and complex process level.
• Stepwise and cross-use case integration approach (monitoring tool and plugfests) is
a pre-requisite for a complex integration such as the ADAS&ME project required.
• Close collaboration/communication between integration teams at Use Case level.
• Enough resources at each partner involved in the integration.
• Proactive reserve of funds (limited) for any support to purchase integration
consumables or software if required.
The work on “Integration” was performed mainly in Work package 6 of ADAS&ME.
3.6 Evaluations
Work package 7 concerned the data collections and evaluation work done for all the seven use
cases. The work involved the creation of the evaluation framework, rre-pilot data collections,
the final evaluations of all use cases carried out. and the consolidation of the results of the use
case evaluations.
3.6.1 Evaluation framework
The Evaluation framework developed in ADAS&ME defines a methodology for evaluating the
system functions based on two principle questions concerning (1) whether the system does what
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 36 of 68 Version 2.1
it is supposed to do within its defined specification and (2) how the users experience the system
regarding their responses and behaviour observed in conjunction with its use (Cocron et al.,
2019). The activity was based upon the development of appropriate data collection protocols
and tools which are applicable to all the use cases with the intention to make it possible to
consolidate the findings from all use cases. The work included detailed test protocol, with
information needed to conduct a successful evaluation. Amongst this were the specific data
collection methods and tools for objective and subjective measures. These included items such
as questionnaires and objective data points from physiological sensors and vehicle performance
measures. Also defined was the methodology and process for driver state inducements required
for the testing of use case demonstrators. Subjective measures of driver state were also specified
to quantify the success / failure of the inducement. By extension fall-back strategies were also
investigated, with these included in the evaluation procedure to account for eventualities where
the inducement procedure. These were then integrated into the general framework defined, but
with individual evaluation strategies adapted to fulfil the requirements of each individual use
case.
3.6.2 Pre-pilot data collections
The pre-pilot data collections were a series of tests with the intention to collect empirical data
regarding the components of each use case required to support the developers of driver/rider
detection systems, but also those working on environmental issues. The testing methodology
comprised a 3-phase process, see Figure 30 which involved test planning as part of the initial
phase. The work was aimed to support the data collection phase 2 and subsequent sharing
between the relevant work package partners.
Figure 30 Overview of 3-phase testing process.
The pre-pilot data collections were conducted to support the algorithm developers, the sensor
developers and the use case teams. All use case teams performed at least one pre-pilot data
collection, but most of them at least two different tests (Anund et al., 2020). The data collections
were carried out across two iterative stages. The first iterative testing stage involved a series of
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 37 of 68 Version 2.1
experiments conducted across the different use cases. The testing focussed on driver states and
their effects. The second iteration of testing was carried out regarding some early
implementation of ADAS&ME system components for all use cases. The data collection and
the results during the pre-pilot are confidential. Data was however always made available to
partners through the sharing platform.
3.6.3 Final evaluations
The final evaluations of the use cases and the related scenarios covered a series of technology
demonstrator platforms with various sensors, algorithms, HMI, and automated control functions
developed as part of ADAS&ME. The evaluations involved the scoping and planning,
development off the experimental procedure, including technical functionality control of the
ADAS&ME systems as well as their effects on real users focusing on driver/rider state,
behaviour and perception. The realisation of the final evaluations and the consolidated results
are possible to read more about in Deliverable 7.2 (Anund et al., 2020).
To ensure that the developed systems can accurately detect when the driver/rider is not
capacitated to drive, and consequently mitigate these states and avoid dangerous situations
using adapted HMI and automation modes, the overall ADAS&ME final evaluation aims were
to:
1. Evaluate the effectiveness of the systems to recognize the driver’s/rider’s state.
2. Assess the capacity of the HMI to display clear and unambiguous information.
3. Evaluate driver/rider behaviour following a system warning/suggestion.
4. Collect the driver’s/rider’s opinion on the system’s usability.
5. Assess the driver’s/rider’s trust and acceptance levels regarding the ADAS&ME
functions.
In total 198 participants (48 of them were female) have been involved in data collections related
to creation of datasets for the development of algorithms and during the final evaluations.
3.6.4 UC A - Truck
The evaluation of use case A was carried out in a Scania truck fitted with the ADAS&ME
system concept shown in Figure 31. The evaluation procedure was modified due to technical
limitations with the demonstrator vehicle (no automated system, truck not able to be driven
with the system active). This led to objective data being collected primarily for manual driving
with the induced driver state, with the intention being to demonstrate the experience of driving
in such a scenario. The evaluation took place on the ADAS test track at the IDIADA proving
ground, see Figure 32. A stationary ADAS&ME system demonstration was then conducted
with a simulated automation system used to present the system functions. In total 7 truck drivers
participated.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 38 of 68 Version 2.1
Figure 31 Interior of Scania test truck (left) stationary demonstration system (right).
Figure 32 ADAS test track at IDIADA
3.6.5 UC B – Electrical passenger car
An open road test route, see Figure 33 - Figure 34 was used to induce anxiety, providing a
highly representative real-world test scenario. The vehicle entered the test track areas at
IDIADA for activation of the automation system. There were issues with the anxiety
inducement due to the test conditions. However, the fall-back strategies were effectively used
to trigger the system functions. A total of 14 participants were tested and the system concept
was generally well received, although there were some apparent issues with comprehension of
the HMI. A positive impact on the acceptance of battery-electric vehicles was shown.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 39 of 68 Version 2.1
Figure 33 Open road test route used for UC B.
Figure 34 Participant driving UC B demonstrator vehicle on the test route with HMI
active.
3.6.6 UC C/D - Conventional passenger car
The evaluation of Use Case C/D was carried out in DLRs test vehicle fitted with the
ADAS&ME system concept, see Figure 35. In total 8 German speaking participants were tested
using a within-subjects design across 2 sessions.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 40 of 68 Version 2.1
Figure 35 Use Case C demonstrator vehicle with integrated ADAS&ME components.
The results show that the system functioned well in some cases, although there were some
scenarios where the ADAS&ME components were not active as planned. The participants
experienced the system positively overall.
3.6.7 UC E/F - Motorbike
A session of extended riding on the ‘General Road’ at IDIADAs proving ground, see Figure
36, was used to induce physical fatigue aiming to closely replicate the physical demands of long
distance riding journeys
Figure 36 General road test route used for UC E physical fatigue inducement.
The distraction component was unfortunately abandoned due to a non-functioning algorithm.
Thus, there was no data collected for visual distraction. A total of 14 participants were tested at
IDIADA, see Figure 37. The results show that the fatigue detection concept was well received,
as was the HMI. However, the performance limitation function was generally seen as negative.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 41 of 68 Version 2.1
Figure 37 Participant with Ducati Multistrada with integrated ADAS&ME components.
The evaluation of use case F was moved to the Ducati test track facilities due to the
developmental nature of the prototype motorcycle used to demonstrate the stabilisation
function. The evaluation was carried out with a small sample of 3 expert riders evaluating the
system performance. For safety reasons it was not possible to make rider faint, that is actually
induce one of the critical states. Hence this state was manually triggered in order to evaluate
the developed stabilisation function.
3.6.8 UC G – Automated docking at bus stop
The evaluation of Use Case G was conducted in a bus simulator at the VTI facilities, see Figure
38. The evaluation was carried out with 16 participants, see Figure 39.
Figure 38 VTI bus simulator used for the Use Case G evaluation.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 42 of 68 Version 2.1
Figure 39 Participant driving the VTI bus simulator in Use Case G.
All data collected have been analyses and a consolidation of results is presented in Deliverable
7.2 (Anund et al., 2020). The data and the results are also used as an input to the impact
assessment in Deliverable 8.1 (Meta, Shingo Usami, Azarko, Jackson, & Verschuur, 2020b)
and also in Deliverable 9.4 covering the exploitation results of ADAS&ME (Verschuur, 2020)
(confidential).
3.6.9 Demanding issues during the final evaluation
The work with data collections and final evaluations had to respond to various events occurring
throughout the project prior to the implementation of the evaluation procedure in the final
evaluations. These events primarily stemmed from issues associated with technological
maturity regarding individual components, algorithms, and the final demonstrator platforms
themselves. The technological maturity issues became especially apparent during realisation of
the final evaluations. Above all, in the evaluation’s changes had to be made from the already
defined experimental procedure in order to respond to integration issues. In some cases,
evaluations had to be postponed and there were issues with missing data due to inconsistently
functioning data loggers.
Unreliability of driver state inducements was also an issue, with these not proving effective in
many cases. This prevented the full assessment of the system functions and resulted in the use
of fallback solutions to activate the ADAS&ME HMI and automation functions.
Test track availability also resulted in some issues and following changes to original proposals
defined in the evaluation framework. This came about for two reasons (1) a delay in the overall
project schedule moving the planned testing outside of the ‘low season’ testing window at the
IDIADA proving ground and causing scheduling conflicts leading to changes and (2) risk
assessment of the evaluation procedure that required changes from the original plan to ensure
the safety of experimenters and participants.
3.6.10 Achievements
Overall the outcomes of the evaluation work package led to the development of a framework
for evaluation of the ADAS&ME system. The framework was successful in providing a generic
template of comparable methods which could be applied across the diverse test vehicles and
locations used for the evaluations. The evaluation framework provided an output that was both
detailed and comparable between use cases. Also, as part of the work package the evaluation
framework was successfully implemented to assess the components of the ADAS&ME system,
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 43 of 68 Version 2.1
whilst being able to respond to required changes.
3.6.11 Recommendations
For future evaluations it is recommended that a process of demonstrator technical integration
is fully completed well in time before the commencement of evaluations is possible.
Recommended time before evaluations are at least more than 2 weeks before to ensure that any
late problems can be responded to and would give a better opportunity to consciously react and
adapt the evaluation protocol accordingly. Perhaps most significantly this would minimise any
impact on the schedule for final preparations.
The work on “Evaluation” was performed mainly in work package 7 of ADAS&ME.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 44 of 68 Version 2.1
4 Use Case Approach and Achievements
In the following section a brief description of each use case in terms of aim, approach,
achievements, innovations and limitations is presented. For more details of the results please
read ADAS&ME Deliverable 7.2 (Anund et al., 2020).
4.1 Use Case A – Truck
As higher automated vehicles become available, the opportunity exists to use automation to
compensate for compromised driver states. However, as the movement of control between
driver and vehicle becomes possible, there is a risk for confusion regarding who (driver or
vehicle) is responsible for what aspects of control at different times. This was one area
addressed within the ADAS&ME project. Furthermore, the design of the system has the
potential to significantly influence trust (and distrust) in the system which can ultimately affect
system usage. Additionally, a failure (on behalf of the driver) to understand the system status
or mode can significantly compromise safety. Within UC A effort was made to define and create
a system that includes a driver monitoring system and an adaptive HMI that works with an
autonomous vehicle.
4.1.1 Aim
Based on the assumption that automated driving, in general but especially during periods of
compromised driver states, will support safer, more efficient and pleasurable road transport UC
A defined four primary goals: 1. Detect periods of compromised driver states.
2. Create an HMI that:
a. Motivates the driver to handover control to the automation.
b. Builds trust between driver and vehicle.
c. Facilitates effective and pleasurable handovers/takeovers.
3. Develop smart automation that reacts to driver state.
4. Detect when during automated driving the driver is resting.
4.1.2 Approach
Consistent with the ADAS&ME work package structure, the process by which the ADAS&ME
system was developed involved the steps described below. The key use cases were identified
based on a user and stakeholder perspective. Where possible, development of the ADAS&ME
system was consistent with this and at the same time considered the general development
direction of technology within Scania and the different goals and timescales that exist between
ADAS&ME and Research and Development at Scania.
With the scenarios defined it was necessary to create a vehicle architecture around which the
ADAS&ME system could be built. The architecture was created in consultation with other UCs
to ensure that as much work as possible was re-usable between use cases, see Figure 40.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 45 of 68 Version 2.1
Figure 40 Summary of UC A architecture.
Based on the defined architecture, efforts were undertaken to define and determine environment
and driver states. Initially, the intention within Use Case A was to make use of an automated
vehicle platform being developed internally within Scania. For the environmental issues, the
effort was oriented towards adapting the environmental monitoring platform of the Scania
platform to be consistent with the ADAS&ME vehicle architecture. This included developing
or adapting the IDIADA test track maps for use with the Scania automated vehicle. Due
primarily to safety constraints, an automated vehicle was not available for use within
ADAS&ME. Consequently, an automation simulator was developed to compensate for this,
and the remaining ADAS&ME system components were created ‘plug-and-play’ ready for an
autonomous truck.
The work on driver impairment detection was focused on specifying what driver states could
be valuable to long-haul truck drivers and how they could be measured. To support the
algorithm developers two separate data collections were done. The first was with 10
professional drivers who drove two 450 km drives while data was collected from a range of
physiology monitoring sensors. A constraint identified post-hoc with this data collection was
that only a few of the desired driver states were occurring naturally. Thus, in the second data
collection a simulated driving task was used where the relevant driver states were induced, see
Figure 41.
Figure 41 Photo of Stage I (left) and Stage II (right) data collection.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 46 of 68 Version 2.1
The data collected supported the algorithm development. Additionally, information collected
influenced the ongoing HMI development and a major contribution was the development of the
HMI strategy which guided the development of the Decision Support Module, see Figure 42.
The HMI Strategy defined the state machine of behaviour related to different driver states,
environmental events and automation modes.
Figure 42 HMI Strategy Overview table.
In addition to the overarching HMI strategy and all the graphical and auditory files (and
programming of their behaviour) for the included HMI elements were created and tested. The
six HMI elements used were Instrument Cluster, Secondary Display, Audio Display, Seat
Vibrators, Steering wheel LEDs and Cabin LEDs, see Figure 43.
Figure 43 Graphical representation of HMI elements (left panel) and photo from
simulator testing (right).
With the architecture defined, the driver monitoring method and algorithm developed and the
HMI elements and behaviour defined the integration into the truck took place, see Figure 31.
4.1.3 Achievements
The results show that the participants spontaneously handed over when automation became
available, both in a non-elevated driver state (normal state) and when distracted, frustrated or
sleepy (impaired driver state). It was only during frustration that two participants did not
handover directly. All drivers seem to understand how the handover and takeover procedures
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 47 of 68 Version 2.1
should be performed. The handover is done simply by pushing a button on the steering wheel,
with no other requirements on the driver, and was done without problem by the drivers during
all driver states. The takeover is a longer process that requires more of the driver but seems to
be equally understandable. Even though all drivers needed at least two tries before successfully
finishing their first takeover, the system was described in positive terms by all drivers. The
drivers found the takeover procedure strict but necessary. Some negative comments by drivers
related to how strict the gaze detection was during takeover. In general, the drivers found the
HMI understandable and clear. They especially liked the use of coloured LEDs on the steering
wheel and the instrument panel and the seat vibration as an escalated warning. The system was
well accepted, and the drivers commented that it would make it safer to drive with an elevated
driver state. Confidence in use, ease of use and system integration were given high ratings on
the System Usability Scale. System acceptance and trust were also very highly rated.
The drivers seem approve of the HMI and warnings for sleepiness and distraction, whereas for
frustration, the comments from the drivers were more mixed. The reason for this could be
related to how clear the risk is perceived for each driver state. Strong warnings were accepted
for sleepiness and distraction, but not for frustration. One recurring comment was that for
sleepiness the seat vibration should be triggered for moderately sleepy as well, and not only for
critically sleepy. The rationale was that it is better to warn the driver in time than wait for the
driver to become critically sleepy. For critically sleepy, several drivers commented that the
system should forcefully take over control.
One common error by the participants was that they did not look forward and did not have both
hands on the steering wheel during the takeover procedure in their first tries to take over control
(when trying the procedure uninitiated during the discoverability test). In the subsequent trials,
all participants completed the takeover. The system was designed to make it easy for the driver
to do a handover when automation was available, and to be strict and check the driver’s fitness
to drive manually when doing a takeover. The results show that all participants completed the
handover at their first try, and all drivers were pushed to look forward and have both hands on
the steering wheel for five seconds to complete the takeover. Some of the negative comments
regarding the strict takeover procedure related to a system error, resulting in gazes directed to
the instrument cluster not always being detected. This speaks for the original concept which
was designed to accommodate for gazes towards the instrument cluster, where the takeover
instructions were shown. Some of the drivers’ comments indicated that gazes directed towards
the mirrors should also be allowed , which could suggest that a more complex distraction
algorithm is needed to allow for the regular gaze behaviour used during driving, and not only
for gazes on the road ahead. Another reason for drivers failing to do the takeover procedure was
that the system did not register that the driver had both hands on the steering wheel while
pressing the steering wheel button. This was likely caused by an overly strict hands-on detection
system that did not allow for the hand movement needed to reach the button.
Four participants removed their hands from the steering wheel directly after automation was
activated when the system prompted that the automation was to be engaged. After 15 seconds,
a confirmation showed that the system was now driving the vehicle and that it was safe to
remove the hands from the steering wheel. This was when the drivers were expected to release
the steering wheel, while their actual behaviour (acting on automation activation) was likely
caused by an unnecessarily long delay before showing the confirmation.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 48 of 68 Version 2.1
Some of the key achievements from UC A include:
• Creation and implementation of a new driver monitoring system and adaptive HMI
that considers and influences automated driving mode.
• Full data flow from sensor data, driver state detection, decision module, and HMI
triggering/Automation change.
• Creation and development of the demonstration truck with open architecture
accessible to all (relevant) partners.
4.1.4 Innovations
Some primary innovations developed in UC A include:
▪ Use of a single sensor (eye tracker) to support detection of four driver states
(sleepiness, distraction, frustration, and resting).
▪ HMI strategy and Adaptive HMI Logic.
▪ Video based emotion algorithm enables continuous tracking of emotional state and is
integrated into a Jetson hardware.
▪ Well accepted and understood Handover and Takeover procedure.
4.1.5 Limitations
There are two primary limitations within Use Case A. First, although the ADAS&ME system
was designed to be ready for use in an autonomous truck, due primarily to safety reasons the
solutions developed were not deployed or tested in an autonomous truck. Despite this, the
developed system listens to and sends output to automation mode even though there is no actual
autonomous system. Related to this is that the system could not be tested in the intended way,
that is, with full automation possible. An additional limitation for the final testing arose due to
safety constraints associated with certain hardware integration, full manual control with all HMI
elements working was not possible. A solution was developed involving full manual control
(limited HMI), simulated manual control (full HMI), and simulated autonomous control (full
HMI) of the ADAS&ME truck.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 49 of 68 Version 2.1
4.2 Use Case B - Electrical vehicle
Range anxiety (or range paradox) is a concept emerged at late 90s which is the concern of not
reaching to the destination or to the next charging spot while traveling in an Electrical vehicle.
This is a stressful experience of a present or anticipated range situation, where the range
resources and personal resources are in fact available to effectively manage the situation,
however they are perceived to be insufficient. Studies show that, electric vehicle drivers usually
need around 160 km of autonomy per charge. Nevertheless, they often prefer vehicles with
considerably higher available range (around 350 km). This demand (which seems to be
avertable) comes from the worry of experiencing such a situation in the future or present, worry
of what happen if such a situation emerges, worry of not being able to find a solution to the
situation and from the worry to being stranded in this uncomfortable situation. If the
manufacturers cannot lower the range anxiety to near zero, electric vehicles will not be able to
compete with gasoline and diesel cars. Acceptability of electrical usage is often linked to the
anxiety of running out of battery power. This UC demonstrates how a better management of
the driving range information, coupled with an anticipative and protective approach, can
decrease this anxiety.
4.2.1 Aim
The aim was to develop a system with an HMI that effectively support drivers to reduce their
level of anxiety and at the same time reach their destination without running out of battery
power.
4.2.2 Approach
The work around this use case followed the same structure as the others considering system
architecture, specifications, driver detection algorithms, HMI development, verifications,
integration and final evaluations.
Two data collections took place before the final evaluation. Both were aimed to collect
empirical data related to electric car drivers’ emotional states while driving. Data was used
mainly to design, develop and test algorithms toward an automatic emotion recognition based
on facial expressions and speech. The data collection was also used to understand how to induce
range Anxiety among the electric car drivers.
The environment, the driving task or even the HMI itself might contribute to anxiety creation.
Hence, it is mandatory to verify whether the perceived anxiety is due to the range. The
experience of range anxiety is assumed to be expressed on the following levels:
1. Cognitive level: (i.e., negative cognition associated with range like concerns about
running out of energy and not being able to reach the destination)
2. Emotional level: (i.e., changes in affect associated with a range situation like feeling
of nervousness or even fear)
3. Behavioural level: Decreasing immediate anxiety by increasing perceptions of safety
and control, (i.e., certain activities like tapping with fingers on the steering wheel,
changing driving style to save energy or frequent checking of relevant displays, e.g.,
range and navigation display or yelling, honking, aggressive gesturing)
4. Physiological level: Under parasympathetic control (i.e., blood pressure, heart rate,
heart rate variability, galvanic skin response (skin conductance), cortisol level, pupil
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 50 of 68 Version 2.1
diameter) and under sympathetic and parasympathetic control (i.e., respiratory
frequency).
In the light of previous information, range anxiety detection is accomplished in two distinct
steps. The first step is detecting basic anxiety by combining the cognitive, emotional and
physiological levels of expression. Although these clues are fundamental to detect the anxiety,
they are not capable to define the actual source of it. Hence, the information from the
behavioural expression level is used to conclude if the detected anxiety is related to the range
or not. The key data was to define if the driver was frequently checking the range information
displayed on the HMI while he was anxious.
4.2.3 Achievements
The results from the final evaluation show that the participants did not fully accept the system.
They rated most items near the zero-reference point. No aspect was rated negatively, but most
aspects only slightly positive which shows that there is some space for improvements. The
participants rated the system as quite pleasant, good and effective. Furthermore, they indicated
that the system is able to raise alertness. In addition, participants found the system quite easy to
use and would imagine that most people would learn to use the system quite quickly. They
assessed their trust on a medium level which means that they were not fully confident with the
system, and they were not totally convinced that the system provides safety. They rated the
system as quite integer, dependable, reliable and trustworthy and felt quite familiar with the
system. This might be due to the design of the data collection were the test leader played a role
convincing the driver that it was true they were soon out of battery power.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 51 of 68 Version 2.1
4.3 Use Case C/D - Conventional passenger car
4.3.1 Aim
The aim was to develop a system that provide a driver state based smooth and safe automation
transition in situations when unsuccessful handover due to non-reacting driving is the case. The
aim is to develop such system both during controlled and emergency situations.
4.3.2 Approach
Use Case C focused on driver state based smooth and safe transitions between manual and
automated driving. Therefore, tailored HMI strategies for the use case specific driver states
were developed and evaluated. Four different driver states were addressed in Use Case C:
sleepiness, stress, frustration and distraction.
When in manual mode, the ADAS&ME system checks if the driver is in an adequate driver
state. If the driver state changes into an inadequate driver state, the HMI reflects this and gives
the driver feedback regarding the detected driver state. Additionally, the ADAS&ME system
offers the opportunity to change to an automated driving mode. If the driver agrees, the switch
to the automation mode is communicated and the vehicle automation is in control of the vehicle.
When the driver experiences automated driving, the automation will constantly check for
possible future situations which cannot be handled by the automation, i.e. environmental
conditions where automated driving is no longer possible. In these situations, the driver
functions as a fall back and needs to take over the vehicle control. If a takeover by the driver is
needed, the system will also consider the driver’s state. In case the driver is capacitated to drive,
the system will perform a standard transition. However, if the driver is in a degraded driver
state (sleepy, stressed, frustrated or distracted), the timing of relevant HMI messages is adapted
to the criticality of the specific driver state, supporting the driver in performing smoother and
safer transitions.
The function developed for use case D is an escalation of use case C. Similar to use case C,
once driving on a highway in SAE level 3, when approaching a system limit a driver state based
transition request (from automated to manual) is sent. However, the driver does not react to the
request and fails to take over control of the vehicle in time. Due to the non-reacting driver the
automation needs to intervene and starts a safe stop manoeuvre. The safe stop manoeuvre results
in lane changes to an emergency lane and braking manoeuvre to a full stop. If there is a lead
vehicle the follow me function will be started. The follow me manoeuvre results in following
the lead vehicle and then in a lane change with a safe stop. These functions were implemented
in the same vehicle as use case C functions, i.e. DLR’s test vehicle.
At first relevant driving scenarios for this use case were identified and agreed on. In a next step,
data collections were conducted to gather more insight in the inducement procedure of the
different driver states and collect sensor data for the algorithm development. Furthermore,
interaction strategies for smooth and safe transitions were developed and assessed in simulator
studies. As a last step the ADAS&ME functionalities were implemented in DLR’s test vehicle
(FASCarII), a modified VW Passat which was equipped with multiple computers and sensors
necessary for driving in highly automated mode (SAE3). The FASCar II was equipped with
project specific sensors and computers from project partners to detect driver states and elements
in the driving environment. Software (automation and environmental sensing) as well as
hardware prototypes were also included. Further, a second braking pedal was installed at the
passenger seat. When the FASCar II was driven in highly automated mode a safety driver was
required in the passenger seat to bring the vehicle to a safe stop in case of an unlikely event of
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 52 of 68 Version 2.1
a system failure. Further information regarding technical and hardware architecture of the
FASCar II can be found in ADAS&ME internal Deliverable D2.1. For the automated functions
in use case D an expert rating was performed to find the most accepted thresholds for the safe
stop manoeuvre, see Figure 44.
Figure 44 The UC C/D Car.
4.3.3 Achievements
The results of the evaluation reveal the potential of the developed driver state adaptive HMI
transition strategies in Use Case C and D. Regarding transitions for control from manual to
highly automated driving, the results show that using the ADAS&ME HMI with driver state
specific HMI strategies reduces the time of manual driving in an inadequate driver state. A key
factor is the “knowledge” of the system regarding the actual driver state. If the driver is in an
inadequate driver state, the automation can prompt the driver to activate the automation earlier.
This strategy results in a shortened time of manual driving in an inadequate driver state.
However, we could not find an effect in reducing take-over times by using the developed
ADAS&ME HMI. The ADAS&ME HMI does not reduce the time to take-over vehicle control
after automated driving compared to a standard HMI without driver state adaptation.
Nevertheless, take-over times in both HMI conditions were very fast (< 6 seconds) so both HMI
variants seam to support the driver in performing the transition of control. The main reason for
not finding differences between the HMI designs could be the arrangement of the scenario. Due
to test track limitations, we could not let participants experience a longer period of pure
automated driving (> 20 min). Further, for safety reasons, there was not a dangerous situation
in front of the car while performing the take-over. Since there was no urge to intervene fast (no
urgency of the situation) participants may have reacted a bit later which led to prolonged take-
over times. A major point for the evaluation of a transition design is the success rate. The
comparison of both HMI strategies shows important differences in these criteria. By using the
ADAS&ME HMI with driver state specific strategies lead to a higher rate in successful
transitions of control compared to the standard HMI. Especially in the distraction trials the
ADAS&ME HMI helps the driver to take over the vehicle control after driving in an automated
mode more successfully than a standard HMI. Again, this illustrated the potential of the
ADAS&ME HMI design to enhance traffic safety. Use Case D evaluated the HMI design if the
driver was not reacting to the take-over request by the system. Participants experienced a
minimum risk manoeuvre by the vehicle ending in a full stop. Participants rated the HMI
solution positively and trustful.
Some of the key achievements in UC C/D included:
▪ Planning and realisation of V2X communication.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 53 of 68 Version 2.1
▪ Planning and realisation of the agreed software architecture.
▪ Development of a driver state specific HMI for transitions of control from highly
automated to manual driving and vice versa.
▪ Development of a tailored interaction strategy for 4 different driver states.
▪ Development and integration of strategies for personalized HMI in demonstrator
vehicle using results from 2 HMI studies.
▪ Realisation and integration of the specific HMI strategies including hardware and
software into the demonstrator vehicle.
▪ Integration of partner Hardware (5 PCs) into physical architecture of one demonstrator
vehicle.
▪ Integration of a project wide software middleware into the demonstrator vehicle.
▪ Integration and realisation of automated driving functionality in demonstrator vehicle.
▪ Planning and realisation of the final evaluation at IDIADAS test track in Barcelona.
4.3.4 Innovations
Some primary innovations developed in UC C/D include:
▪ Development of well accepted and safe driver state adapted interaction strategy for
vehicle HMI.
▪ Tailored interaction timing for criticality of driver state and environment.
▪ Personalisation of interaction strategies.
▪ Integration of hardware and software for the assessment of 4 different driver states
simultaneously in one demonstrator vehicle.
▪ Realisation of driver state assessment for automated driving.
▪ Connected automated functions for driver state related minimum risk manoeuvre in
case of a non-reacting driver.
4.3.5 Limitations
Even though the evaluation study on a test track provided important insight into the
effectiveness of the developed system there is still a lack of understanding of its effectiveness
on the real road. In addition, the evaluation study suffered from practical problems during the
testing which resulted in a lack of objective data on system effectiveness. More detailed
research is needed for investigating the developed HMI strategies in different road types and in
different (also urgent) scenarios. Further, the driver state specific HMI strategies for the
transitions need to be adjusted continuously to provide results in this research area and the
development of driver state assessment needs to continue. Not all sensors used in the prototype
system are practically useful for the automotive industry or in line with legal conventions (e.g.
large computer in the trunk for processing data, microphones in driver’s field of view).
4.4 Use Case E/F - Motorbike and protective gear
During long-range motorbike touring, environmental conditions (e.g. extremes of temperature),
combined with rider fatigue, can affect the rider’s physiological (e.g. resulting in rider
dehydration) and psychological state and lead to high-risk situations. Further in situations where
the rider is losing consciousness the motorcycle becomes uncontrollable resulting in serious
accidents. For the above situations the rider states of physical fatigue, inattention and stress
were studied.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 54 of 68 Version 2.1
4.4.1 Aim
The aim was to develop a system that assist the rider during long range touring, in case of
physical fatigue and inattention. In case the rider faint the system should also activate a support
system including automated stabilization of the bike and speed reduction.
4.4.2 Approach
The function developed for long-range attentive touring in use case E supports riders during
long-range motorbike touring. Based on environmental information and rider state, the system
detects signs of physical fatigue, distraction or stress and sends a warning to suggest the rider
to make a pause. The detection is based on sensors integrated in the protective gear together
with some of the HMI solutions used for warning and information the rider, see Figure 45.
Figure 45 Rider protective gear with integrated sensors.
If the rider does not comply, a recovery mode is activated limiting the motorcycle performance
without stopping it. The function limits the power and, consequently, the motorbike’s
longitudinal acceleration in order to not exceed a predefined speed. To deactivate this mode
and have the full motorcycle potential again, the rider must take a break or to power off the
motorbike.
For the development of the detections system a data collection was required to understand the
development of rider states, especially physical fatigue, inattention and stress. Those data were
collected both in simulators and on real road.
The final evaluation of the function used for UC E was tested at IDIADA’s proving ground and
surroundings by using a Ducati Multistrada 1260 equipped with the rider state detection
algorithms and HMI as well as other communication systems. The stabilisation system to be
used during rider faint in use case F was evaluated at DUCATI due to safety issues and time
constraints.
4.4.3 Achievements
The main results for the long-range touring and rider faint perspective are the fact that an
innovative capsize control algorithm capable of stabilizing a motorbike was designed. The
system also works in the presence of unbalanced loads or errors affecting the roll angle
estimation. In addition, an HMI devices manager, tailored to get the best feedback combination
to inform the rider, were developed and integrated. Also, the integration and interaction with a
wearable electronic sensor and unit outside the back protector is an innovation to be mentioned.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 55 of 68 Version 2.1
From a technical perspective, a full set of wearable electronic devices with a wireless
connection, allowing rider physiological state monitoring and visual, acoustic and haptic
feedback, was developed. The HMI devices aimed to offer effective rider support in case of
fatigue and distraction and the results show a close to production solution. From a safety
perspective, a set of functions to mitigate high-risk conditions such as torque limitation and
capsize control was developed, but also a system that uses an automated HMI strategy for
smooth transitions and effective rider warnings.
The main achievements in UC E/F are:
▪ Design of innovative capsize control software capable of stabilizing the motorbike even
in presence of unbalanced loads or errors affecting the roll angle estimation
▪ Integrated and coherent HMI devices management, tailored to get the best feedback
combination to inform the rider
▪ First integration of wearable electronic sensor and unit outside the back protector
4.4.4 Limitations
A clear limitation is the fact that a new system like the once develop is very safety critical to
evaluate with real riders. Only professional test riders were allowed to drive and to safety
aspects since rider faint and rider distraction manipulation was not possible to do due to safety
reason.
4.5 Use Case G - City bus
Driver fatigue has received increased attention during recent years and is now considered to be
a major contributor to approximately 15–30% of all crashes. The main cause of driver fatigue
is sleepiness due to sleep loss, being awake for too long, and driving during the circadian low.
Also, work-related factors such as stress and shift work contribute to driver fatigue. In addition,
it is important to consider the type of task, as both cognitive underload and overload contribute
to demanding situations influencing the drivers. The goal of doubling the public transportation
travels by 2020 requires more efficient operation, and already now working as a bus driver
involves a lot more than just driving the vehicle. The responsibilities to control where to go,
keep track of the timetable, make sure that the bus is on time, oversee and support ticketing,
communicate with the operator and interact with the passengers can be overwhelming. On top
of that the bus driver occupation is associated with negative physical and psychosocial factors
related to driver’s health. The factors described as most important are poor in-vehicle
ergonomics, risk of injury, shift work, working alone, time pressure, manoeuvring within
increasingly congested urban environments, risk of intimidation and violence, accessibility
problems as well as poor health, in particular obesity. Many of these factors are also expected
to become more severe in the future and lead to an even more stressful work environment. High
levels of work-related stress and disturbed sleep is a dangerous combination contributing to
diseases and poor workplace performance. One possible way to support bus drivers is by
introducing automated functionalities that release them form some tasks, to avoid overload and
at the same time improve the comfort and safety for passengers and people outside the bus.
Such a functionality might be automated docking at a bus stop.
4.5.1 Aim
The aim of this use case was to develop and evaluate the effect of a future system using
automation during docking at bus stop.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 56 of 68 Version 2.1
4.5.2 Approach
The automated docking function was system initiated and the drivers were asked to give control
to the bus when approaching the bus stop, but also to take back control when the bus stop zone
was coming to its end. When leaving the bus stop zone the system decided if the drivers were
fit to get the control back, taking into consideration detection of sleepiness and distraction but
also confirming that the drivers had at least one hand on the steering wheel, see Figure 46.
The development of the HMI strategy for the involved scenarios included three different steps.
The first step was an exploratory study with the aim to understand bus drivers’ working
conditions and problems. This was done during an experiment on real road with 15 city bus
drivers involved. The second step was a virtual reality (VR) simulation study with 10 bus
drivers in which a first HMI concept was evaluated. The outcome of the VR study was then
modified and integrated in a moving-base driving simulator at VTI, Sweden. The third step was
a pre-pilot with 7 city bus drivers driving both in an alert and an expected sleepy condition.
Tuning and adjustment of the HMI, detection algorithms and automated functionalities were
made based on the outcome of the pre-pilot.
The final evaluation took place using an urban city bus route scenario with in total 20 bus stops
(10 with and 10 without automated docking available). In total 16 participants drove twice, once
in an alert and once in a sleepy condition. The design was a within-subject design and the order
of driver state was balanced. The same simulator as in the pre-pilot was used.
Figure 46 Use Case G scenario with HMI for an automated docking functionality
integrated
4.5.3 Achievements
It can be concluded that the results from the final evaluation of the ADAS&ME system with
automated docking at bus stop showed that automation did not result in a significant difference
in self-reported stress and blink duration. There was however a significant increase in KSS and
in HRV (RMSSD) indicating lower arousal. In addition, there was a significant reduction in
speed.
In general, there was an effect of what the driver experienced first, manual driving or
automation. The drivers starting with manual driving showed higher KSS and longer blink
duration during automation. The highest levels on sleepiness indicators were seen during
automation with expected sleepy drivers, supporting an effect of time on task. For those starting
with automation there was also an effect of automation on blink duration and HRV indicating
more sleepiness signs and less arousal during automation. Looking into driving behaviour, it
could be noticed that departing from the bus stop showed slightly lower accelerations with the
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 57 of 68 Version 2.1
automated system activated. A large increase in the amount of lost tracking and glances towards
“other” was revealed when the bus was about to take over control from the driver. The drivers
liked the system, wanted it and thought it could contribute to increased safety.
4.5.4 Innovations
Some primary innovations developed in UC G include: ▪ A system for automated docking at bus stop.
▪ An integration of automation and driver state detection that only gives back control if the
driver is fit to drive defined by being alert and/or attentive and has a least one hand on the
steering wheel.
▪ An HMI strategy for transition that is easy to understand and to use.
▪ An automated function that reduces the speed while docking (deceleration and acceleration).
4.5.5 Limitations
There were differences in the results depending on if the driver started to drive with or without
automation. It seems that there is a carry-over effect and an interaction effect if the driver was
expected to be alert or sleepy. The results clearly indicated that the system first has one effect
which is changing with time on task. How to avoid this needs further investigation.
The evaluation of automated docking at bus stop suffered from some limitations. One is that
the system was not evaluated in a real road environment. Therefore, it is difficult to say if the
results are valid also for real road situations. This is something that needs further investigation.
To be able to understand the effect of the developed system it is essential to implement it in a
bus on real road and look at real stress levels and on the user’s opinion of the HMI. In the
conducted study the focus was on the drivers, but in future research also the passengers’
perspective is relevant to include. It is important to understand how to communicate that and
how the bus is docking automatically to the passengers to guarantee trust and acceptance among
users.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 58 of 68 Version 2.1
5 Assessment of the impact
The assessment of the impact has been undertaken from a multicriteria perspective which
reflects the macro areas of impact: Safety, mobility and environment, economic and social,
Legal and regulatory issues. For more details see Deliverable 8.1 Impact assessment (Meta,
Shingo Usami, Azarko, Jackson, & Verschuur, 2020a).
Except for the Legal area, for which the analysis is conducted across the UCs, the impact
assessment has been undertaken for each use case separately considering safety, mobility and
environmental impact as well as the socio economic one. The methodology applied for the
assessment cover mainly the following:
▪ Safety, mobility and environmental impact assessment ▪ Analysis of user-related aspects (from drivers and potential users’ point of view) ▪ Cost-Benefit Analysis (CBA) ▪ Stakeholders’ analysis
5.1 Safety, mobility and environment
The impact assessment is closely related to several parts of the project as described in Figure
47.
Figure 47 Impact assessment and its relations to other parts of the work in ADAS&ME
ADAS&ME has its priority on safety and therefore also the safety impact is of high priority.
The work with safety impact followed the procedure described in Figure 48.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 59 of 68 Version 2.1
Figure 48 Safety impact assessment procedure
From the safety impact point of view, it can be said that the ADAD&ME systems meet the
principal aim of counteracting the underlying causes behind risks that affect the safety of all
road users, contributing to increased safety for both the driver and passengers of vehicles fitted
with the ADAS&ME system, and other road users that come into contact with these vehicles.
There is evidence from all the use cases supporting the success of the system in mitigating
effects of adverse driver states. Thus, it is possible to conclude that the ADAS&ME systems
can contribute to the prevention of resultant accidents that cite these errors as route causes.
Moreover, it is possible to say that there is an improvement in lateral and longitudinal
accelerations that is seen through integration of the system on buses (UC G). In addition,
evidence is found to support the possibility of heightened trust and acceptance associated with
the ADAS&ME systems. In some cases, this can be viewed as a positive impact – improving
the likelihood of users using the ADAS&ME system such that the other safety benefits can
manifest. There is, however, evidence found within the literature to suggest a link with reliance-
associated behaviour which could have the potential to increase risks of driver error in some
situations.
Concerning the mobility impact, with the evident reduction in aggregate journey time
associated with use of the ADAS&ME system in UC G, an immediate improvement to mobility
of passengers can be supported. This is in conjunction with the evidence for reduced journey
time having a positive effect on mobility.
For environment, data collected in the ADAS&ME evaluations generally supports the
hypothesis of improved efficiency following its implementation. It can be said there will be a
positive impact on emissions following increased battery electric vehicle use. Also tested were
potential increases in efficiency brought about by influences on driving behaviour.
5.2 Economic and social impact
The socio-economic impact assessment has it starting point in the user related aspects. The
results show a positive an overall attitude regarding the acceptance of ADAS&ME systems
among the respondents (drivers who experienced the system and survey respondents). The most
desirable Use Case is UC C/D which has the high rate for willingness to use as well as for the
trust and willingness to pay extra money. The cross-analysis among Use Cases shows that all
systems were rated positively but there were differences between the Use Cases. The systems
developed in use case A, use case C/D as well as use case G were assessed as somewhat positive
1. Definition of target accident
population
2. % of target accident in
EU28
3. Forecast of target accident
in 2030
4. Market penetration
5. Safetybenefits
estimation
6. Cost – unit rates
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 60 of 68 Version 2.1
with a tendency to very positive. The system developed in use case B and part of the system
developed in use case E/F were assessed as somewhat positive with a tendency to neutral.
Overall, the CBA has shown different results according to each UC. Cost-Benefit Ratio (CBR)
values in low and medium scenario are lower than 1, showing a socio-economic inefficiency.
The main reason for these poor results is the fact that the costs of the systems are quite high,
especially for UC A and UC C-D. The other use casess show a CBR greater than 1 in the high
scenario. This means that the target year 2030 may be too early to reach any significant
penetration rate and, thus, benefits for all the systems.
The results of the Stakeholders Analysis have shown that Authorities and Policy makers were
the most skeptical among the Use Cases as well. The main barriers to the full scale deployment
of such systems reported were mostly related to the acceptance and trust aspects, confusion
about HMI operations, increase of a complexity of the systems which lead to doubts in
reliability and safety of the vehicle equipped with ADAS&ME systems.
5.3 Legal and regulatory impact assessment
The Legal assessment has highlighted that the General Data Protection Regulation (GDPR)
does not impede ADAS&ME but imposes several significant obligations. In particular, the
commercialisation of certain personal (such as health data) may be incompatible with the
GDPR. However, no specific solutions required at this stage.
Concerning Traffic Rules compliance, most of the national systems assessed in the present
impact study currently allow for completely autonomous driving (SAE levels 4 and 5), i.e.
without a driver being able to control the car (within the car or remotely). Legislative change
seems necessary to allow for (temporary) control by automated driving systems. In particular,
the definitions of 'control' and 'guardian' may have to be changed in various (national) legal
systems. Finally, regarding Insurance and liability issues, the existing notions of 'control',
relevant for attributing liability, might not be adequate for automated driving. Specific rules
should be considered in respect of (i) the allocation of liability when vehicles are in automated
driving mode and (ii) the burden of proof for consumers in case of deficiencies of automated
driving systems.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 61 of 68 Version 2.1
6 Exploitation of the results
The main goal of this activity was the creation and validation of the business model concepts
of ADAS&ME innovations, including the Use Cases. A total of 26 innovations were selected
for the realisation of the safety, economic and environmental potential of integrated technical
solutions and were further analysed, see Table 4.
Table 4 Exploitable results of ADAS&ME
Exploitable Results UCs Partner(s)
New tools for customer insights via unobtrusive monitoring A SCANIA, DENSO
Sustainable product development via VR based HMI development
tools
A SCANIA, DENSO
Human Controller state modules as individual components A SCANIA, DENSO
Resting module for haulage company A SCANIA, DENSO
External HMI for incapacitate driver A SCANIA, DENSO
Automatic calling of 112 A SCANIA, DENSO
HMI personalization package A SCANIA, DENSO
Smart route planning for individual drivers based on physiological state A SCANIA, DENSO
EV range anxiety B VALEO, VEDECOM
Driver impairing emotions detection system C EPFL
Environmental situation awareness module C TOMTOM, DENSO
Heart rate and respiration detection steering wheel radar1 C VALEO
Camera based sleepiness and inattention detection system2 C SEYE
Stress and discomfort detection seat sensor C FORD
HMI and automated function steering wheel3 C AUTOLIV
Driver state-based smooth and safe automation transitions C DENSO
Non-reacting driver emergency manoeuvre function D DLR, DENSO,
TomTom, FORD
Long range attentive touring function with motorbike E DUCATI, DAINESE
Rider faint function F DUCATI, DAINESE
Passenger pick-up/drop-off automation function for buses G VTI, SCANIA
Passenger pick up/drop off automation for buses G VTI
HMI steering wheel with LED bar and 2-zone hands on detection
sensor
G AUTOLIV
Driver identification module B, C, D SEYE
Driver personalisation module A, B, C, D, FORTH
1 Previously referred to as: "Inattention/distraction detection steering wheel sensor". 2 Previously referred to as: "Camera based drowsiness and inattention detection system". 3 Previously referred to as: "HMI and automated function modules".
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 62 of 68 Version 2.1
G
Rider identification module E, F DUCATI, DAINESE
Rider personalisation module E, F DUCATI, DAINESE
The following conclusions were reached:
▪ As to IP rights, some of the innovations are patentable, while some others are not
(mainly due to existing patents). In some cases, the technology developed will be
publicly disseminated so that all market players can further develop and exploit the
innovations.
▪ The expected market prices differ substantially (depending on the nature of the products
and the markets involved), but for most products they are expected to amount to several
hundreds of EUR.
▪ The market potential is generally seen as (relatively) positive for most products.
▪ With the exception of a limited number of products (which are expected to be ready for
the market within the next 12-24 months), the time market for many products is at least
3-5 years and for some even 5-10 years. A number of products are still in the
development phase.
▪ The market risks identified mainly concern a perceived low added value of the products
(in the eyes of the potential users), the novelty of the products (and the associated risk
of a perceived lack of reliability), health concerns (electromagnetic waves) and a lack
of standardization/interoperability. Furthermore, the Legal Impact Assessment (part of
WP8) identified potential legal impediments in the area of Data Privacy (GDPR),
Traffic Rules, and Insurance and Liability.
▪ The exploitation routes differ in function of the state of development of the products
(many of which require further research and/or development) and the intentions of the
consortium partners involved (for example, research institutions do not have the
intention to exploit the innovations on the market, while some of the partners intend to
use the innovations for internal purposes rather than externally).
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 63 of 68 Version 2.1
7 Dissemination
Communication actions and dissemination activities have been of high importance from start
to end of the ADAS&ME project. In order to achieve high impact of ADAS&ME results several
offline and online tools were developed, highlighting the project website, social media networks
and videos tailored for all the audience. In addition, dissemination of results at key events,
collaborations with other EU-projects, and writing scientific papers with ADAS&ME partners
but also with externa collaborators have been important. The exploitation of the results obtained
during the project lifetime was also a key aspect for the project and its legacy, a market analysis
was carried out to support the business plan of the ADAS&ME innovations. Firstly, a wide
socio-economic impact analysis was created through a CBA (cost-benefit analyses) considering
the stakeholder consultation and finally the Business and Exploitation plans for 17 potential
ADAS&ME developments. All dissemination activities are describe in Deliverable 9.7
(Figuals, 2020).
The dissemination plan and activities were developed at the beginning of the project and then
updated every year. Definition of our audience and the creation of the project image with
logotype, roll-up and all the promotional aspects including the use cases’ illustrations took
place. The website was developed at the very beginning of the project and reached 50,000
webpage visits during the project lifetime well supported by the social media networks (Twitter,
LinkedIn and Youtube channels) that exceed the KPIs included at the dissemination planning
with more than 1,000 followers. One of the most important tools to promote the project and
explain the progress was the videos, three animated videos explaining the ADAS&ME concept,
the motorbike use cases and the car use cases were produced with the objective to create
awareness to a wide audience. At the end of the project the use case demonstration videos were
developed to secure a legacy of the developments achieved during the project, Figure 49.
Figure 49 ADAS&ME project Youtube channel.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 64 of 68 Version 2.1
A list of key stakeholders was created. Those stakeholders were invited to the Stakeholder
Workshop set in Brussels where the selection of the ADAS&ME use cases and the
corresponding use case scenarios along with their impact were defined. This first contact was
successful for the project development and it built a close contact with several external
stakeholders who have been used for promoting the project through social media, increasing
the performance of user survey used for the impact assessment and inviting participants to the
Final Event, see Figure 50.
For videos from the final event please visit the following links:
• Final Event Use Case Demonstration: https://www.youtube.com/watch?v=qpQvBG8XNwM
• Final Event promo: https://www.youtube.com/watch?v=rGFgjb55nOo
• Interview Anna Anund: https://www.youtube.com/watch?v=A4EkxFDorfU
• Interview with Angelos Bekiars: https://www.youtube.com/watch?v=O8xIQMo9TiU
Figure 50 Participants at the ADAS&ME Final event December 2019.
The most relevant aspect of dissemination was the large participation of ADAS&ME in national
and international events/conferences. The total was 35 direct participations including the most
relevant of the European framework: EUCAR Conference (Brussels), stand at the EUCAR
Conference (Brussels), special sessions at ITS Europe Congress (Copenhagen and Brainport),
the co-organization of Autonomous Vehicle & Development Symposium (Stuttgart) and
participation in INEA stand at TRA (Vienna) among other international workshops and events.
In addition, 25 scientific papers have been published during the project lifetime, see also
Deliverable 9.7 (Figuals, 2020).
Throughout the project ADAS&ME partners have reached out and shared knowledge with the
EU funded projects I-DREAMS, PROSPECT, AutoMATE, MAVEN, TrustVehicle, VI-DAS,
TransAID and BRAVE. As an example, ADAS&ME co-organized with AutoMATE the
AutomotiveUI (Oldenburg) and Intuitive Partially and Highly Automated Driving Conference
(Aachen). In addition, this Activity was also extended to collaboration with other initiatives
such as CardioID Technologies which gave ADAS&ME interesting inputs during the project
lifetime and resulted in a conference proceedings titled “Driver drowsiness detection: a
comparison between intrusive and non-intrusive signal acquisition methods”. Finally, during
the last stage of the project ADAS&ME were in contact with the new awarded Automated Road
Transport projects Suaave, Trustonomy and Drive2theFuture in order to share the knowledge
generated by the project.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 65 of 68 Version 2.1
The market and socio-economic analyses had an objective of demonstrating the economic
viability in the current uncertain society and it paved the way for the future exploitation. Some
market challenges were detected and analysed in other activities of the project (data and
security, legal issues and regulation and ethics). Fully analysed challenges were GPS
technology, safety, infrastructure connectivity and, especially important, economic aspects and
adoption of end users. The sources of information for the market and socio-economic analyses
were ERTRAC, JRC and other EU funded projects. The analyses showed valuable conclusions.
In economic terms, it is expected that Cooperative, Connected and Automated Mobility
provides profitable opportunities for sectors like automotive, electronics and software,
telecommunication, data services, digital media and freight transport. Sectors like insurance,
maintenance and repair are identified as businesses that might suffer important decreases in
revenues in the future, especially as a result of decreasing numbers of accidents. At a societal
level, a Cooperative, Connected and Automated Mobility could bring important safety and
productivity gains. Nevertheless, some important concerns exist, such as users’ acceptance,
ethics, social inclusion, and labour. ICT competences will be increasingly demanded in the
future, e.g. in manufacturing, maintenance and transport-related jobs. The skills required for
driving a vehicle will also change as automation gains full control of the vehicle, e.g. requiring
more supervision and selective skills.
During the project lifetime changes of data regulation at European level have appeared (i.e.
GDPR) and a Question & Answer has been developed including all knowledge gathered during
the project and useful information from project partners. As a conclusion of this activity, the
following key aspects were developed:
▪ GDPR took effect and additional considerations were added to data principles.
▪ Special focus on sensitive data: the knowledge of what is personal data is a key for an
effective business and exploitation plan.
▪ There are many differences between pilots and exploitation in the field of automated
vehicles.
The main objective of the ADAS&ME Advisory Board has been to supervise the project results
and key outcomes and support the dissemination of knowledge generated by the project. The
Board members are Jim Sayer (UMTRI), Ashleigh Filtness (Loughborough University), Olivier
Lenz (FIA Region I) and Masao Nagai (JARI). At the beginning of the project the Advisory
Board supported the selection of Use Cases at the Stakeholder workshop organized by
ADAS&ME. In addition, they reviewed and provided expert feedback on the project mid-term
results and the development of the systems. Finally, they validated the final project results at
the final project demonstration event in Barcelona.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 66 of 68 Version 2.1
8 Conclusions
ADAS&ME aimed to develop an Advanced Driver Assistance Systems that incorporate
driver/rider state, situational/environmental context, and adaptive interaction to automatically
transfer control between vehicle and driver/rider and thus ensure safer and more efficient road
usage for all vehicle types (conventional and electric car, truck, bus, motorcycle).
The work was built around seven use cases using in total five demonstrators. A generic system
architecture and HMI strategy were used and adapted to each use case. Driver/rider state
detection algorithms useful not only when driving manual but also during automation were
developed. The detection system was then the trigger for adapting the HMI solutions needed
for both manual driving and transitions (both driver/rider and system initiated). The knowledge
of the environment added information that made especially the takeover accurate and safe. The
seven use cases were evaluated, and driver state was manipulated if safety was possible to
guarantee. If not possible still the HMI was evaluated. The results where then used to assess
the impact and to identify new business and possible exploitation of results.
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 67 of 68 Version 2.1
9 Contacts
9.1 Coordination team
Coordinator
Dr Anna Anund, VTI
Email: [email protected]
Technical manager
Dr Evangelos Bekiaris, CERTH/HIT
Email: [email protected]
Innovation manager
Stella Nikolaou, CERTH/HIT
Email: [email protected]
Quality manager
Dr Lena Nilsson, VTI
Email: [email protected]
9.2 Dissemination manager
Marc Figuls, RACC
Email: [email protected]
9.3 WP leaders
WP1 Dr Anna Anund, VTI Email: [email protected]
WP2 Sri Venkata Naga Phanindra Akula, TUC Email: [email protected]
WP3 Dr Karel Kreuter, DENSO Email: [email protected]
WP4 Dr Mathissen, Marcel (M.), FORD Email: [email protected]
WP5 Dr Frederik Diederichs, IAO Email: [email protected]
WP6 Stella Nikolaou, CERTH/HIT Email: [email protected]
WP7 James Jackson, IDIADA Email: [email protected]
WP8 Dr Eleonora Meta, CTL Email: [email protected]
WP9 Marc Figuls, RACC Email: [email protected]
WP10 Dr Anna Anund, VTI Email: [email protected]
9.4 Use Case leaders
UCA – Truck: Dr Stas Krupenia, SCANIA Email: [email protected]
UCB – Electrical vehicle: Kevin Nguyen, MOVEOTEC Email: [email protected]
UCC/D – Conventional vehicle: Marc Wilbrink, DLR Email: [email protected]
UCE/F – Motorbike: Davide Sette, DUCATI Email: [email protected]
UCG – Bus: Dr Anna Anund, VTI Email: [email protected]
ADAS&ME (688900) Deliverabel 10.4
February 2020 Page 68 of 68 Version 2.1
10 References
Anund, A., Solis, I., Rauh, N., Jackson, J., Ahlström, C., Abser, A., . . . Georgoulas, G. (2020).
Use ase Test results and consolidation (Deliverable 7.2 ADAS&ME EU project).
Retrieved from
Cocron, M., Vallejo, A., Delgado, B., Wilbrink, M., Anund, A., Krupenia, S., . . . Harous, C.
(2019). Evaluation frame work - ADAS&ME project (Deliverable 7.3 ADAS&ME EU
project). Retrieved from
Diederichs, F., Knauss, A., Wilbrink, M., Lilis, Y., hrysochoou, E., Anund, A., & Krupenia, S.
(2018). Adaptive Transitions for Automation in Cars, Trucks, Busses and Motorcycles.
IET Intelligent Transport Systems.
Dukic Willstrand, T., Anund, A., Strand, N., Nikolaou, S., Touliou, K., Gemou, M., & Faller ,
F. (2017). Driver/Rider models, Use Cases and implementation scenarios (D1.2
ADAS&ME EU project). Retrieved from
Figuals, M. (2020). Report on project dissemination and international cooperation activities –
2nd Update ADAS&ME (Delivearble 9.7 EU Project ADAS&ME). Retrieved from
Hennes, N., & Mathissen, M. (2020). Deliverable 4.2 Driver/rider state detection module
(Deliverabel 4.2 ADAS&ME EU project). Retrieved from
Meta, E., Shingo Usami, D., Azarko, A., Jackson, J., & Verschuur, S. (2020a). Deliverable 8.1
Impact Assessemnt (Deliverabel 8.1 ADAS&ME EU project). Retrieved from
Meta, E., Shingo Usami, D., Azarko, A., Jackson, J., & Verschuur, S. (2020b). Impact
assesement ADAS&ME (Deliverabel 8.1 ADAS&ME EU project). Retrieved from
Touliou, K., Maglavera, M., & Britsas, C. (2017). SoA and Benchmarking ADAS&ME (D1.1
ADAS&ME EU project). Retrieved from
Verschuur, S. (2020). Deliverable 9.4 Buisness and Exploitation Plan (Delivearble 9.4
ADAS&ME EU project). Retrieved from