5 - 7 June 2018
Messe Stuttgart, Germany

Conference Programme

Day 1: Tuesday 5 June

Room A Keynote Opening Session
09:00 - 12:45

Test development and execution system for autonomously performed scenarios

Mathias Klumpf
Test & development engineer
Audi AG
To ensure highest quality and safety, complex testing of autonomous driving assistance is performed with every vehicle during SOP and later on at random. To include a maximum number of sensors and actors, autonomous parking procedures are chosen as test scenarios. The action of the vehicle is observed by a micro GPS system. The quality is identified by a complex fuzzy-logic-based algorithm automatically.

Road attribute confidence values using HD maps

Michael Laur
Senior technologist automated driving
The presentation will discuss a method of adding probabilistic confidence values to look-ahead road topology and world model attributes using HD maps.

Insuring autonomous (test) driving

Andreas Bradt
Project manager autonomous driving
Allianz Automotive
After more 130 years of automotive development, the autonomous car is almost here. Autonomous mobility will exist in many shapes, and insurance companies need to prepare for self-driving and the new risk factors. During the transition, the transfer of responsibility must be designed with safety as a high priority. Insurance products and processes will need to be adjusted to account for autonomous driving. Allianz has already gathered on-road experience by insuring autonomous vehicles globally. Not every accident can be prevented by an ADAS, and a standardised data storage system in automated vehicles is required.

Validation of L4 urban automated driving

Dr Christian Krummel
Robert Bosch GmbH
Validating a fully automated vehicle operating in urban scenarios is currently the ultimate challenge. Statistical approaches are neither feasible nor comprehensive. The talk will address current views of challenges and approaches in releasing self-driving vehicles.

Room A Test, Verification & Validation
14:15 - 18:00

Map-based validation of autonomous vehicles

Dr Henning Lategahn
Atlatec GmbH
Autonomous driving will hit the road in the near future. One of the obstacles that is still to be overcome is the validation of this technology. How can one prove that the vehicle senses the environment correctly? Two approaches are currently followed. One is real-world driving tests and the other is virtual testing in simulators. Both seem tempting. The first, however, suffers from the unavailability of accurate ground truth data, whereas the latter suffers from unrealistically simplified virtual worlds. We present an approach to acquire accurate 3D map data that can be used as ground truth and realistic 3D world models.

Experiences gathered from the application of the ENABLE-S3 V&V architecture

Dr Andrea Leitner
Research project manager ADAS
AVL List GmbH
This presentation aims to give an overview of first results of ENABLE-S3, a large European project with around 70 partners from industry and research. The project’s goal is the provision of technology bricks (methods, tools and models) and a generic test architecture to enable more efficient testing of automated systems. Standardisation is a major enabler for modularisation and reuse. Therefore, the consortium evaluates and extends existing specifications such as OpenScenario and OpenSimulationInterface. The talk will highlight some first results showing advantages and limitations identified in selected demonstrators of the project.

Autonomous drive controllers challenge validation and verification test data management

Alexander Noack
Head of automotive electronics
b-plus GmbH
Centralised computing platforms for autonomous driving functions – the domain controllers – require a huge amount of input data. Multiple cognitive sensor sources in autonomous cars challenge the whole testing and validation scenario. Huge computing power processes gigabit data that forms the basis for decisions. Conventional ECU development methods such as prototyping platforms, vehicle loggers and simulation systems such as hardware in the loop are reaching technological limits. The presentation focuses on the changes in sensor architecture and the challenges of methodology for measurement data management in the domain controller environment.

Maximising test coverage utilising high-fidelity vehicle record and playback techniques

Nicholas Keel
Group manager, automotive product management
National Instruments
As autonomous functionality in vehicles becomes more sophisticated, the test coverage required continues to expand. Traditional techniques such as hardware-in-the-loop testing provide an excellent foundation upon which to establish an autonomous vehicle test infrastructure, and track testing provides real-world sensor data required to thoroughly exercise the entire active safety system from sensor to software. However, system-level testing often requires sensor fidelity that is difficult to capture in simulation, and extensive track testing is incredibly expensive. By combining real-world logging with lab-based HIL simulation, you can generate the highest-fidelity data streams and play them back in a variety of controlled scenarios to maximise test coverage.

Unified system of tools for ADAS

Dr Florian Baumann
Technical director
Adasens Automotive GmbH
We are proposing a unified system of tools to efficiently test and validate computer vision algorithms. This includes the categorisation of different self-developed tools for pre-development, development and post-development stages. The pre-development stage consists of recording and planning tools. The development stage consists of tools to visualise output of cameras, lidar or radar. The post-development stage is to test KPIs such as ROC curves, false positives, etc. This talk will inspire people to develop such tools internally. All tools are running via a web front-end, and demos will be shown during the presentation.

Systems approach to creating 'interesting' test scenarios for autonomous vehicles

Siddartha Khastgir
PhD researcher
WMG, University of Warwick, UK
To prove that automated driving systems are safer than human drivers, it is suggested that they will need to be driven for over 11 billion miles. However, rather than number of miles it is the quality of miles that is important. To find the 'interesting scenarios', a novel systems engineering method – an extension to the STPA method – has been developed and applied to low-speed automated driving systems (Pods). One of the features of the proposed method is the ability to create both test scenarios and pass criteria.

Room B Vision - Sensors, LiDAR and Mapping Technology
14:15 - 18:00

Maps As a Sensor For The Autonomous Car

Mike Tzamaloukas
Vice president, BU autonomous drive/ADAS and CoC navigation
With the advent of the autonomous car, modeling the environment has been a challenge and an opportunity of unprecedented scale. Mike Tzamaloukas, Vice President, BU Autonomous Drive, ADAS and Navigation, HARMAN will present a novel framework that ensures that every connected car on the road has the ability to benefit from validated, crowdsourced information. In this presentation, Mike will address topics spanning efficient multi-sensor data collection, normalization and machine learning cross-validation, spatio-temporal confidence estimation, data sharing and customer privacy. He will further offer insights on how this technology can help reduce production quality costs and accelerate go-to market timelines for OEMs, fleet operators and transportation city planners.

NDS and OADF map standards for successful AV deployment

Dr Volker Sasse
VP Navinfo / chairman NDS and OADF
Navinfo & NDS
This presentation will address NDS & OADF map standards, which are key for successful deployment of autonomous vehicles. History shows the map was always of high importance, however, AV’s require the support of better maps. Therefore, the whole world will need be mapped extremely precisely. For example; Japan already collected all freeway maps for autonomous driving at Olympia 2020. But AV’s require worldwide standard maps and more cooperation, as each car acts on map data collection, other updates are too slow. Up and down streams will fill the air with map data. Within OADF, several standards like NDS, ADASIS, SENSORIS and TISA will cooperate.

Enabling autonomous vehicles in inclement weather

Phil Magney
Founder and principal
VSI Labs
Enabling automated driving in inclement weather is challenging because sensor performance is compromised. For example, a light snowfall that covers lane markings would render most methods of lane keeping inoperable. However, there are new methods for localisation that heighten the performance of automated vehicles. In this presentation we will share our experience using precision map data to improve the AV performance in poor weather. By using different methods of localisation against the precision map, VSI will discuss how it uses HD maps to improve the performance and safety of automated vehicles, even when lane markings are covered or absent.

Integration of solid-state lidar in vehicles: best practices for superior object detection

Filip Geuens
Technology choices for automotive lidar must be based on how and where the lidar units can be integrated in vehicles. This integration impacts the lidar technology as well as the vehicle itself. As the need for reliable solid state lidar systems continues to grow, considerations about sensor positioning and ways to achieve reliable detection are gaining automotive attention. This presentation will report on some of the outcomes from the research cooperation with Tier 1 partners for specific and integrated lidar positioning. The use of lidar as a reference or complementary system for camera-based or sensor-fused detection will also be addressed with regard to the lidar vs. camera positioning.

Determining precise boresight alignment of a lidar for autonomous driving

Richard Sands
Application engineer
Oxford Technical Solutions Ltd
This presentation covers a recent case study that addresses the challenges faced determining precise alignment of lidars in an inertial navigation system frame. We have developed a reliable calibration method to remove systematic errors caused by mechanical misalignment. We will share techniques used to validate the calibration result, along with specific findings that will help to improve confidence in open-road sensor testing and validation projects.

Challenges of Multi-sensor and Fusion –ECU measurement for ADAS level 3-5

Alfred Kless
Business development manager
Vector Informatik
The presentation shows the challenges and solutions for a complete ADAS logging approach: multichannel vehicle bus logging: CAN-FD, FlexRay, LIN, Auto-Ethernet; radar measurement technology for raw data as well as object data; high-end fusion ECU measurement with multi uC + uP architecture based on Autosar and Autosar adaptive operating systems; multiple vehicle and reference camera measurement; other sensors such as laser scanner, ultrasonic and GPS. The unique solution covers two use cases: 'Engineer Mode' with full graphical object overlay display, and 'Taxi Driver mode'.

Room C Testing in the Urban Environment
14:15 - 18:30

Complete approach for testing automated vehicles on a testing ground

Dr Houssem Abdellatif
Global head autonomous driving and ADAS
The physical testing of highly automated vehicles is a very challenging task. Highly accurate and repeatable replay of complex manoeuvres must be achieved on the test ground to judge the ability of autonomous vehicles to cope with different situations. For this purpose, we present a complete technical setup that has been developed by TÜV SÜD and its partners. Soft static targets, remotely controlled traffic simulation vehicles, a vehicle control system, highly sophisticated sensors and a monitoring system work together precisely. Technical details and demonstrations are provided in this talk.

Using the StreetWise scenario base for virtual safety assessment

Sytze Kalisvaart
Project manager integrated vehicle safety
For virtual safety assessment of automated driving vehicles, a set of test scenarios is needed with real-world validity. The StreetWise scenario database is based on real-world driving data. Parameterised observed variants of the scenarios are stored. When using the database for virtual testing, the test engineer will have to select a relevant set of scenarios for the system under test. A data-driven approach for determining the relevance of the scenario is proposed. Also, the interface ‘scenario database – simulator tool’ and relationship to key performance indicators is presented.

London's Smart Mobility Living Lab

Iwan Parry
Head of connected and autonomous vehicles
This presentation will provide a review of London's Smart Mobility Living Lab, part of the UK Government's £100m investment in creating an integrated CAV testbed ecosystem for on- and off-road testing. The Smart Mobility Living Lab will build on established and ongoing CAV projects in London, including GATEway and MOVE_UK, and provide infrastructure to support real-world testing, development and evaluation of CAVs and CAV mobility solutions together with new ITS and communications technologies.

Importance of real-life deployment of autonomous vehicles in mobility services

Raphael Gindrat
The democratisation of autonomous vehicles will be directly impacted by the quality of the resulting mobility services and their integration in the current transportation infrastructure. In June 2016, Switzerland was one of the first countries to set up a service of electric autonomous shuttles circulating through a city centre in pedestrian areas and on open roads. Today, deployments of autonomous shuttles are blossoming worldwide, and are crucial to address challenges inherent in public acceptance and the integration of autonomous vehicles in urban areas. The presentation will offer a deep dive into existing autonomous vehicle deployments around the world.

Exploring automated bus lines and cooperative driving

Dr Arie P van den Beukel
Assistant professor
University of Twente
Reliable automation within complex traffic situations, especially in cities, is challenging. For city transport, municipalities face high costs of bus lines. These seem inefficient because buses are used intensively during rush-hours but at other times by only a handful of people. Nonetheless, municipalities need to offer basic transport facilities. Because the main expense is labour costs, some municipalities are considering automated buses within their towns. To optimise availability, costs and reliable operation, a system design is being explored with small bus units that individually drive autonomously in dedicated lanes and drive cooperatively with a human lead driver in town.

Panel Discussion - Get the maximum from real world testing

Dr Houssem Abdellatif
Global head autonomous driving and ADAS
Sytze Kalisvaart
Project manager integrated vehicle safety
Raphael Gindrat
High quality data generation, public perception, repeatable real-world scenarios - we discuss how to get the most from your real-world testing programmes.

Day 2: Wednesday 6 June

Room A Validation in the Virtual Domain
09:00 - 12:45

Training and validating automated driving applications using physics-based sensor simulation

Robin van der Made
Product manager software and services
TASS International (a Siemens Company)
One of the latest needs in the area of automated driving is the generation of sensor data as input for deep neural networks for the purpose of training automated driving applications. The PreScan simulation platform can be used to generate virtual sensor data of all sensor technologies relevant to automated driving, such as camera, radar, lidar, ultrasone and DSRC. In this presentation we also present the added value of injecting synthetic sensor data directly into platforms such as the Mentor Graphics DRS360 and Nvidia Drive PX2 for virtual validation of automated driving applications by means of hardware-in-the-loop (HiL) simulation.

Virtual twin of the London CAV testbed

Jon Horsley
Programme director
Digital Engineering & Test Centre
The UK Government has recently funded a coordinated set of physical CAV test facilities between London and Birmingham to provide CAV technology developers from across the world with the facilities to develop and test the essential underlying technology and actual vehicles on public roads and safe off-highway environments. Alongside the physical, we are building the virtual testing environments and bringing innovative new technology from gaming, AI, etc. into automotive testing for the first time to achieve the required test quality at much reduced time and cost. This paper will outline this exciting coordinated activity and examine the digital London space.

Object detection and classification based on virtually trained neural networks

Ronnie Dessort
Simulation consultant
In autonomously driven vehicles, object detection and fusion of sensor information enables the vehicle to perceive its environment and decide for certain driving manoeuvres. The virtual development of such systems allows the comfortable definition and reproducibility of a large number of traffic scenarios. In this contribution, a publicly available state-of-the-art algorithm based on deep learning is trained and tested in a virtual 3D world. Special focus is put on variation of different harsh environmental conditions such as rain or soiled traffic signs. The results show that the presented development method can be used as an appropriate complement to common development.

New sensor simulation concept for virtual test drives

Caius Cioran
Application engineer
The market introduction of autonomous vehicles is on the horizon. A crucial element of these systems is the reliable environmental perception by means of camera, radar and lidar sensors. Functions for autonomous driving have to be test-driven hundreds of millions of kilometres. This is possible using virtual test drives and SIL/HIL simulation only. In this context, appropriate sensor models and integration options of sensor ECUs with HIL test benches are the key. This presentation outlines a new sensor simulation concept and a generic interface unit to insert raw data behind sensor front-ends. A unique radar sensor model is introduced.

Traffic simulation for connected and autonomous vehicle technology testing

Dr Alex Gerodimos
Executive director
Connected and autonomous vehicles (CV/AV) have the potential to tackle challenging problems related to safety and efficiency. Accidents, congestion and pollution in urban areas could be substantially reduced if these technologies were adopted on a large scale. One of the main challenges for transport agencies and technology developers alike is to find ways to evaluate the performance and impact of such technologies before they are deployed. Simulation provides a comprehensive, repeatable, cost-effective and near-limitlessly scalable testing environment. As such, simulation provides an excellent digital complement to field testing for regression and pre-deployment scenarios.

Panel Discussion - Harnessing the power of simulation

Robin van der Made
Product manager software and services
TASS International (a Siemens Company)
Jon Horsley
Programme director
Digital Engineering & Test Centre
Caius Cioran
Application engineer
Undoubtedly, simulation is fundamental to validating autonomous driving application. In this discussion, we will reach further in to virtual environment, SIL/HIL and virtual data creation.

Room A Robust Test, Verification & Validation Methodolgy
14:15 - 18:00

Processes, tools and philosophy to develop an autonomous electric truck

Joachim Fritzson
As a new way of thinking about road transport, Einride was facing a lot of challenges. New powertrain technologies had to be merged with autonomous driving logic and new sensor technologies. This required the freedom to choose processes and tools without having to compromise with legacy development, but at the same time facing the pain of starting to work from a blank sheet. This presentation discusses the learning curve up to now, and how ViCANdo has helped in accelerating the process and balance risk, and cut development and validation time by being multi-platform capable.

The structure of test specifications of highly automated driving functions

Dr Hardi Hungar
Team leader verification and validation methods
German Aerospace Center (DLR)
Currently, scenarios are considered as main elements in test specifications of highly automated driving functions. A scenario captures a large number of concrete test cases in a formal way. It must permit these test cases to be generated automatically. And the collection of scenarios must cover all behaviour that needs to be tested. The scenario language needs sophisticated constructs to cover the complex, reactive patterns of road traffic. And the test specification must allow overlapping of cases to be systematically coped with. The presentation proposes a conceptual solution to these problems of formalisation and evaluation.

Challenges for testing with platform robots at high speeds (+100km/h)

Markus Schmidl
DSD Testing
Test speed demands for active safety system testing are continuously rising. Current requests from the automotive industry are up to 130km/h. UFO currently reaches as first platform test speeds of 100km/h. The presentation shows the challenges for the test equipment and the proving ground of testing at these high speeds.

Verification techniques for safety and security in autonomous vehicle software

Dr Mike Bartley
Test and Verification Solutions Ltd
The Innovate UK-funded CAPRI project brings together an experienced consortium of partners from industry, academia and local authorities, working together to deliver a complete end-to-end POD mobility service. The consortium aims to collate sufficient evidence from the deployment trials and simulation testing to support PODs becoming a recognised vehicle classification for use on public roads. This presentation will discuss: verifying and validating the safety and security of the next generation of PODs for the on- and off-road environments, and collating sufficient evidence to support PODs as a new vehicle classification.

Applying proven methods for quantifying test results and test coverage

Rainer Straschill
ADAS/AD strategist
FEV Europe GmbH
The presentation will show that questions of test coverage for self-learning systems can be solved: any self-learning system in our scope, including those with non-deterministic portions, can be identically transformed so that the aspects of self-learning and non-determinism can be considered identically to the problem of the unknown use case population. Next, the presentation will explain how established and proven approaches from different domains (experimental physics, mathematical statistics, data analysis) can be applied to achieve a measure for test coverage in the situation of populations where only basic knowledge about their properties exists.

Testing the connected vehicle

Rosario Trapero
Manager Connected Car Competence Center
The connected vehicle landscape is complex, with different communication protocols and standards involved. Successfully implementing and verifying the variety of safety applications in the connected car requires testing at several different levels, including the physical layer, the protocols and the algorithmic functionality of the applications. In this paper, a comprehensive approach for testing all these different layers is presented, covering laboratory and in-the-field testing procedures.

Room B Open-Road & Real-World Testing
09:00 - 12:45

Testing autonomous vehicles in harsh winter conditions – key findings

Harri Santamala
Sensible 4 Ltd
Sensible 4, a startup focusing on automation technologies under harsh conditions and varying environments, will take its self-driving vehicle to northern Lapland for testing during the winter of 2017-2018 as part of the Aurora Arctic Challenge project. Perception, remote control and multi-sensor redundancy and solutions will be studied in open-road tests. The presentation will describe field-test arrangements, technology, key findings and results obtained.

Real-life testing under experimental clauses – recent regulatory developments

Dr Alexander Duisberg
Bird & Bird LLP
Sandbox testing under real-life conditions only works if the regulatory framework strikes a good balance between incentivising innovation and warranting the safety of others. Federal government is aware of the challenges to enable quick and efficient administrative processes, which may touch on all sorts of regulations. 'Experimental clauses' and related approval processes are making their way into the framework. The presentation analyses examples around autonomous drive and other areas of testing from a legal and regulatory perspective, and how government is pushing ahead to enable faster and better administrative processes to facilitate such 'real laboratories'.

The UK's CAV testing ecosystem

Michael Talbot
Head of industrial strategy
CCAV (UK Government's Centre for Connected and Autonomous Vehicles)
CCAV presented an overview of the UK's new CAV testing ecosystem programme at the AV Symposium in 2017. Since then we've announced the winners of the first £51m competition (matched by industry to £102m) and launched MERIDIAN, the 'CAV hub' that will coordinate the UK's capabilities and promote them internationally.

Real-world driving scenario identification for AV functional safety

Gildas Thiolon
Engineer data scientist
Knowledge of the real-world driving environment involving manually driven cars is fundamental, allowing detailed identification of safety issues encountered by autonomous cars in similar complex situations. The MOOVE project has been created to study this problem. This presentation: explains the context and the developed methodology applied to the collected data; presents the data modelling applied to build the database; identifies the role of crucial variables interfering in the extracted real-world driving scenarios. Thus, improved robustness will be achieved for the autonomous vehicle architectures and systems being developed at the Vedecom Institute.

Controlled field-testing of automated and connected driving in urban environments

Rico Auerswald
Research fellow
Fraunhofer Institute for Transportation and Infrastructure Systems IVI
Current test and validation procedures for automated driving functions use field tests on public roads primarily to identify unknown critical scenarios. These scenarios are then validated in simulations or on test sites. The complex traffic environments required to validate automated driving functions in urban traffic may not be investigated with sufficient detail outside of field tests. Therefore, we present an approach that increases the reliability of field tests by generating scenarios considering real urban roads with public traffic, focused resource coordination for the test procedure and cooperation with connection infrastructure in the Digital Testbed Dresden.

Regulatory testing of autonomous vehicles for trials in residential areas

Niels de Boer
Programme director, CETRAN
Nanyang Technological University
Singapore is embarking on a roadmap towards deployment of autonomous vehicles as a part of the transport system in residential areas. To extend the trial of AVs from the existing testbed area to residential areas, a 'milestone 2' test needs to be passed, after which the licence to test AVs will be amended and the permission to test in defined residential areas will be added. This presentation will give an overview of the test requirements, the test definition and technical assessment methodologies, and the work to improve and extend the testing based on feedback from trials of autonomous vehicles in residential areas.

Room B Using Simulation to Advance System Design & Validation
14:15 - 18:00

Real challenges for simulation in verification and validation of AVs

Dr Roberto Ponticelli
Chief engineer - intelligent mobility
Horiba MIRA
The ever-increasing use of virtual design, verification and validation (vDV&V) tools is supported by a sound base of successful case studies across most engineering and scientific areas (e.g. medicine, aerospace and automotive). Nevertheless, there exists a risk of a false sense of safety when the tools used on vDV&V of safety-related systems are not well understood, particularly in highly complex systems like autonomous vehicles. This talk will address how the modelling methods, physical data and correlation, design and test traceability and coverage, operational boundaries, dynamic environments, hardware in the loop (HIL), scenarios annotation and generation, and other key parameters shape the challenge for AV vDV&V.

Enabling the massive simulation that autonomous driving validation requires

Enguerrand Prioux
ADAS/AV product line manager
Proper SAE Level 4-5 vehicle validation will require – at the least – millions of scenarios to be checked. For project timing and testing conditions reasons, this will be entirely feasible only through simulation. To validate sensor technology and design, computer vision and sensor fusion, the decisional and executive driving agent, and actuation, most companies will have to couple several simulation models (AI, controls, world, sensors, vehicle physics) and perform a massive simulation activity to sufficiently sample the scenarios space. In this presentation we will communicate Siemens' progress in this area through a concrete simulation case.

Key simulation features for autonomous vehicle validation

Thomas Nguyen
R&D projects and AD/ADAS product manager
AV Simulation
Recent projects in France and Germany show that simulation will play a key role in demonstrating the safety of autonomous vehicle systems. The key features for these purposes are scenario and sensor models. At AV Simulation, a new joint venture between Oktal and Renault, we have developed a versatile simulation engine dedicated to virtual evaluation and validation of automotive systems. In this presentation, we will present key integrated features: a powerful generation engine that enables a complete representative set of scenarios, based on templates, and a complete tool chain for sensor simulation, including phenomenological, statistical and full-physical modelling.

Purpose-driven scenario generation

Dr Andreas Höfer
Product manager simulation software
IPG Automotive GmbH
ADAS have undergone a rapid evolution. Formerly integrated as standalone components, modern ADAS are networked with each other, which causes new challenges for development and testing. This calls for new testing techniques, such as scenario-based simulation. There are three basic ways to create scenarios: outlining on the drawing board and manual creation with a scenario editor; generation based on map data with optional enhancement in a scenario editor; generation based on real-world measurements with optional enhancement in a scenario editor. Each approach has its own strengths and weaknesses, which are described in this presentation.

Massively parallel simulation for testing and validating autonomous vehicles

Christopher Hoyle
Technical director
Over the last two years we have developed a scalable, fast, open simulation environment for autonomous vehicle testing and validation. Multiple ego vehicles may share the same simulated real-world environment, each with multiple sensor model feeds. The entire environment may run offline, for automated regression testing, or may run real-time, which also allows multiple human drivers to join the simulation, from driving simulators, adding stochastic, unpredictable, error-prone behaviour, i.e. real human behaviour, into the simulation. The same simulation environment may be shared by multiple OEMs or Tier 1s to evaluate how their autonomous vehicles behave when they interact.

Virtual testing by coupling system simulation with SUMO traffic flow simulation

Dr Jakob Kaths
Product manager
Increasing automation in the automotive sector will lead to a stronger interaction of individual vehicles with surrounding motorised and non-motorised traffic and with traffic infrastructure. This creates new requirements for virtual development and testing that can be met by integrating automotive and traffic engineering simulation tools. To this aim, an approach is presented to seamlessly couple high-fidelity models based on Matlab Simulink and the microscopic traffic flow simulation SUMO. This is achieved by establishing a co-simulation setup between the two tools while using common map information based on the OpenDRIVE road description format and including driving dynamics, controllers and in-vehicle sensors such as camera, radar, lidar and ultrasonic.

Room C Connected Vehicle & Infrastructure Technology
09:00 - 12:45

Intelligent traffic lights: from testing to deployment

Jeannet van Arum
Director smart mobility
Province Noord-Holland
Harm Jan Mostert
Senior advisor smart mobility
Province Noord-Holland
The province of Noord-Holland is working on V2X communication. By putting in place intelligent traffic lights, which are able to communicate either through wi-fi P or 3G/4G, use cases can be carried out. Industry and government identified ‘Day One’ C-ITS use cases as the first promising ones for deployment. A couple of these Day One use cases relate directly to traffic lights. In the last two years the province of Noord-Holland (regional road authority in the Netherlands) combined pilots and deployments for these use cases. This presentation will focus on the lessons learned.

A high-integrity, automotive-grade antenna system optimised for V2X DSRC communications

Dr Oliver Leisten
Technical director
Helix Technologies Ltd
The safety-critical role of V2X DSRC in driver-assist and autonomous vehicle systems requires particular attention to 'real-world' channel/propagation challenges which to-date have not been successfully addressed by incumbent antenna technologies. Helix Technologies is developing an antenna based on dielectric-loaded, multi-filar helix antenna technology that is optimised for use in the V2X DSRC IEEE 802.11p environment and employs the use of a diversity system architecture. The V2X DSRC antenna to be developed by Helix Technologies will employ two ceramic-based, compact, dielectric-loaded, multi-filar antenna elements co-located so as to provide optimum antenna diversity, promote link reliability and eliminate coverage nulls.

Aurora Borealis Intelligent Corridor for snowtonomous driving and CAD testing

Reija Viinanen
Director Aurora collaboration
Finnish Transport Agency
Aurora Borealis Intelligent Corridor is a digital cross-border corridor on E8 from Norway to Finland. The main objective is to enable C-ITS and CAD testing in arctic conditions. The road has been instrumented with supporting infrastructure to promote car makers' CAD trials on public roads. E8 provides an opportunity to test cross-border ITS solutions and snowtonomous driving in arctic conditions, and research what kinds of infrastructure and roads will be needed in the future in terms of automated vehicles.

Solutions for V2X test and validation for the connected car

Axel Meinen
Technical sales manager
S.E.A. Datentechnik GmbH
This presentation will focus on necessary tools for test and validation of upcoming mobile communication like 802.11p or cellularV2X (LTE-V) for connected autonomous driving vehicles. S.E.A. has developed a test platform to test and validate the technologies and the corresponding hardware like ECUs and onboard units with respect to RF communication and compliance, protocol analysis and validation, as well as complex scenario generation including GNSS simulation for open-loop or closed-loop HIL tests. A brief overview of the different test aspects will be given, and a way to analyse and visualise the V2X test data will be demonstrated.

Infrastructure for connected and automated driving in Hungary

Adam Nagy
Traffic engineer
Hungarian Public Road
Future mobility requires more and more online information even from the road. The Hungarian Public Road company is committed to supporting this development. C-ITS deployment started in 2015 in Hungary to demonstrate the use of C-ITS to exchange data through wireless communication technologies between vehicles and infrastructure (V2I). Road safety improvement, especially work zone safety, was the key issue, but the upgrading of services is ongoing. A 136km-long stretch of motorway M1 has been covered, and an extension is planned in 2018 with urban-interurban use cases. The company is also participating in the deployment phase of an automated proving ground and test zone in Zalaegerszeg.

Leveraging V2I communications to enable fully autonomous operations

Chad Mack
Product manager
Global Traffic Technologies
For nearly 50 years, Global Traffic Technologies’ traffic signal priority control technology has been connecting vehicles to infrastructure. This technology has evolved from IR to GPS-enabled radio and cellular-based communications. The V2I communication infrastructure that enables driver-assisted priority control can be extrapolated to autonomous vehicles as technologies advance. GTT’s latest application in New York City leverages existing connectivity on buses and at intersections to create a centralised transit signal priority (TSP) solution. This centralised solution demonstrates how cities and companies can use connected hardware to create a reliable network that is capable of supporting autonomous vehicles.

Supporting vulnerable road users in connected and automated driving landscapes

Dr Fatih Özel
Project manager and consultant
Oecon Products and Services GmbH
Connected and automated vehicles (CAVs) are the subject of extensive research nowadays owing to their potential to improve safety. The Federal Ministry of Transport and Digital Infrastructure in Germany has funded the Digitaler Knoten 4.0 (Digital Intersections) project, which aims to develop and test CAV technologies for traffic intersections. As part of the project, a smartphone app is developed for integrating cyclists and pedestrians (vulnerable road users or VRUs) into the overall cooperative CAV system. This paper presents the existing technological approaches for integrating VRUs in cooperative CAV systems, our adopted approach as well as future development and implementation steps.

Active pedestrian safety testing in CARISSMA

Dr Igor Doric
Senior researcher and deputy scientific and technical manager
The presentation will discuss the results of the pedestrian safety research projects TARGETS and VISYTE. The projects ran from October 2013 to October 2015, and from November 2015 to January 2018 respectively, and were supported by the Federal Ministry of Economic Affairs and Energy, on the basis of a decision by the German Bundestag. Furthermore, the presentation will provide details about the CARISSMA research and test centre, including the CARISSMA rain simulation.

Pedestrian in the Loop

Michael Hartmann
Research engineer
Virtual Vehicle Research Center
A large number of testing procedures have been developed to ensure vehicle safety in common and extreme driving situations. However, these conventional testing procedures are insufficient for testing autonomous vehicles. They have to handle unexpected scenarios with the same risk a human driver would take – or less risk. Currently, safety-related systems are not adequately tested, e.g. in collision avoidance scenarios with pedestrians. Examples are the change of pedestrian behaviour caused by interaction, environmental influences and personal aspects, which cannot be tested in real environments. It is proposed to use a test environment with flying drones. Name: Pedestrian in the Loop.

Day 3: Thursday 7 June

Room A Using a Test Facility to Advance Testing Programmes
09:00 - 10:30

Introducing the UK’s controlled urban testbed for connected and autonomous

Peter Stoker
Chief engineer - vehicle
Millbrook Proving Ground Ltd
The critical need for a semi-controlled but realistic urban test environment for CAVs that seamlessly connects with open-road urban environments, raises unique challenges. Millbrook, in collaboration with the United Kingdom Atomic Energy Authority's RACE, is working to address these challenges. Based on a step change enhancement to existing roads and test capability, the facility will revolutionise preparation and validation of CAVs, services and technologies for public road deployment. It will offer open access to users – including developers of software, sensors, roadside units, telecommunications (5G) and cybersecurity systems – to explore public and industry impacts and accelerate the development of CAV technologies.

Proving Ground Zala – a unique test environment for future mobility

Zoltán Hamar
Technical director
Automotive Proving Ground Zala Ltd
The objective of the project is to establish an automotive testing site for the automotive and communication industry, which would be built on the domestic R&D capacity of the sector and on European R&D capacities equally. In line with current and future automotive trends, when establishing the testing site we place special emphasis on the testing needs of autonomous vehicle systems and related environmental conditions. Proving Ground Zala is a unique test site where the fusion of classic dynamic test elements and test elements of future technologies are realised on a 260ha area.

Autonomous highway driving in undetermined weather conditions

Dr Sergey Shadrin
Scientific advisor
Technology provides autonomous driving within a road lane without visual recognition of the road marking, while reaching speeds up to 130 km/h in undetermined weather conditions (rain, snow, etc.), which is achieved by receiving high-precision navigational data. The composition of the technical solution includes in-vehicle automation kit, hybrid navigation system, safety subsystem of risk assessment. The proposed approach provides highway autonomous driving and supports lane-keeping assistance systems in cases of road marking visual recognition failures. Practical issues will be presented.

Room A Using Simulation for Test Scenario Validation
11:15 - 16:15

Digital simulation is key to autonomous driving safety

Jean-Paul Roux
Senior vice president, EU operations
Exa Corporation
Digital anti-soiling simulation work is particularly significant in the upcoming world of sensor-controlled fully autonomous vehicles. We explain how PowerFLOW software incorporates dirt, dust and water into simulation – with the real-world accuracy that can only be achieved using a transient solver. We explain how the particle-handling capability of the software enables automotive and truck manufacturers to deliver a safer and much-improved driving experience through better control of soil and water accumulation on autonomous driving sensors.

Dynamic ground truth measurement for simulation and post-processing

Steffen Metzner
Technology scout ADAS, simulation and control
AVL List GmbH
For development of ADAS and AD functionalities, use of environment simulation tools is a well-accepted methodology. Most expensive and time consuming is the manual definition and design of complex scenarios as input to the experiments. The DGT (Dynamic Ground Truth) research project investigates methodologies and measurement systems for the capturing of dynamic objects surrounding moving test vehicles without the need for invasive changes (e.g. to mount sensors). The recorded data should be usable for onboard processing (enabling direct evaluation and comparison with the environment model of the vehicle) and offline post-processing for more accurate and detailed data analysis.

Scenario-based performance assessment framework for automated vehicles

Dr Jeroen Ploeg
Principal scientist
In view of recent developments in autonomous vehicles (AVs), the need arises for an efficient AV road-approval procedure. To this end, a safety assessment framework that employs virtual assessment of traffic scenarios is proposed. This framework consists of four components: data acquisition, scenario extraction and parametrisation, virtual safety validation, and physical safety validation. Due to the simulation-orientated nature of this framework, quantitative and statistically relevant measures for safety-related AV performance are obtained while minimising the number of physical tests, thus realising an efficient procedure. The assessment framework is currently being developed in Singapore to accelerate AV deployment.

High-performance driver-in-the-loop simulator for autonomous vehicles

Andras Kemeny
Scientific director
For safe autonomous vehicle products for all, validation on billions of kilometres is needed, thus using both massive and driver-in-the-loop simulation, with a large number of driving scenarios. One of the main critical scenarios is the handover between manual and autonomous mode in different traffic situations. The new high-performance ROADS dynamic driving simulator, commissioned to AVS, will help Renault test these cases in the most efficient and realistic way and design the best experience for the final client as well as providing help for general driver acceptance and public approval.

Simulation, development and test tool

Ali Maleki
Executive director
Nexteer Automotive
The presentation will introduce a simulation, development and test environment tool to log radar and camera data, replay through software in the loop, simulate radar/lidar data at low target or high track levels. The system has the ability to extend this environment by using additional plug-in tools.

Decision-making evaluation by oracle vision

Dr Rémi Régnier
Researcher in evaluation of data processing systems
The SVA Project aims to meet the challenge posed by the complexity of demonstrating the safety of autonomous vehicles through the use of digital simulation. One of the objectives is to provide builders and suppliers with a methodology and simulation tools for the validation of safe autonomous vehicles. We propose a new methodology based on the concept of oracle, a mathematical tool adapted to the ADASS evaluation. Indeed, oracle created an ideal reference to compare with the answer of a system. The presentation will explain how to create this oracle.

Virtual automotive testbeds for automated, reproducible tests

Jörn Thieling
Research assistant
Institute for Man-Machine Interaction at RWTH Aachen University
Today, autonomous vehicles are mainly verified in real-world test drives. These tests are not only time-consuming and expensive but also insufficient for a trustworthy validation, since critical situations rarely occur. To enable efficient system tests, we propose the use of virtual testbeds as a software environment in which 'digital twins' of real vehicles, sensors and environments are simulated (from RWTH). That simulated data is compared with real-world data by using established algorithms for environment interpretation from the automotive sector (with ADASENS input and evaluation).

Room B Test Scenario, Verification and Validation Studies
09:00 - 16:15

ACU internal fault detection logic evaluation method using semiconductor modelling

Byoungmoo Kwon
Research engineer
Hyundai Autron Co Ltd
The objective of our study is to present an evaluation method through semiconductor modelling of peripheral sensors, which efficiently evaluates the ACU’s internal fault detection software logic. This study introduces two different methods to generate internal fault conditions, one by editing a pre-recorded data stream of normal operation, and another by modelling the fault-occurring sectors and adding fault logic blocks to the sectors. We evaluated a test MCU’s software with 386 cases of 95 possible internal fault types and achieved a test time reduction of 70% compared with the current method of internal fault testing.

Secure testing of ADAS functions using a VRX driving simulator

Günther Hasna
Director strategic projects
Optis GmbH
Optis is developing a simulator software for dynamic ADB headlamp functions, named VRX-Headlamp. Due to physics-based simulation algorithms it is possible to replace real night test drives with virtual simulations. As the ADAS function of the light-assist system is dependent on sensor inputs controlling the headlamp, it was also necessary to develop virtual physics-based sensors as cameras and lidar. For safety ADAS functions like the autonomous emergency braking AEB, it is now possible to do a virtual validation of mandatory NCAP ratings like the pedestrian AEB night tests with the VRX-Sensors software.

It’s all corner cases: teaching computers to drive safely

Bruno Fernandez Ruiz
Co-founder and CTO
As the industry is charging towards self-driving perception in all-terrain, all-weather and all-lighting, it remains a challenge to collect and tag enough data to allow for reliable driving algorithms. Applying user self-annotation of driving data is the only way we can truly advance the autonomous car industry and infrastructure. In this session, I will share our concept and methods in creating an autonomous learning 'flywheel' of collect-annotate-learn road data network.

A case study of Level 4 test and validation in Korea

Younggi Song
SpringCloud Inc
This presentation will cover self-driving cars and their test and validation procedure. The system consists of multiple sensors and a controller for target vehicles (conventional vehicle and EV). Based on test scenario L4, each task consisted of a 'start, driving, stop and parking' procedure within the test ground. The service point is from ground IoT sensors and vehicle movement information and will provide task performance as defined. The data from SDC is 5 type of dataset and will be continuously monitored by platform service architecture.

Autonomous vehicle mode-based testing: generate test cases for 3D simulator

Fabrice Trollet
MaTeLo product manager
Based on a Renault-Nissan methodology and other R&D projects, this presentation explains how to deal with the huge combinatorics met in ADAS simulation and how to compute related autonomous vehicle expected behaviour, to generate tests cases. The speaker will present the MaTeLo Model-Based Testing process, in which weather, road infrastructure, other motorist behaviour, EGO system, road events, etc. are designed. He will also explain how to design determinist test cases with great variability. From this MaTeLo model, the speaker will show the automatic generation of thousands of scenery samples using Gibbs sampler algorithms, to automatically build Oktal SCANeR scripts for 3D simulation run.

Chasing critical situations in large parameter spaces

Dr Mugur Tatar
Managing director
QTronic GmbH
It is commonly agreed that testing and validating highly automated driving functions will involve a mixture of tests in real and simulated environments. Even after a functional decomposition in several classes of base scenarios there always exist many parameters with a huge number of possible values. Testing all combinations of values is impossible. In the presentation we discuss possible approaches and challenges in exploring the large parameter spaces in the search for hidden faults and critical situations.

Assurance of autonomous vehicles with authentic data recordings

Bernhard Kockoth
Technology scout
Autonomy driving Levels 3-5 are based on a growing number of safety-related systems that must be secured with millions of kilometres. All vehicle bus communications and raw data from sensors, cameras, lidar and radar, as well as status data like weather and actual maps, must be recorded, authentically. An 8hr test drive easily produces 4TBytes of data, if not 20 to 100Tbytes. The data then must be fed to data centres without causing long pauses in vehicle testing. The enormous amount of data becomes a challenge for measurement equipment in automotive environments. New concepts and solutions are presented to meet sophisticated requirements.

Meeting the high demand for automated tests – the TestCase Generator

Tobias Weimer
Specialist for system and software development
MicroNova AG
The number of tests required for autonomous and electric vehicles has increased dramatically compared with established standards in the automotive industry. This demand can only be handled by means of automation and adapting existing processes. Therefore, we developed a new solution for the test automation software EXAM, called the TestCase Generator. Its purpose is to generate test cases automatically from a test specification. Now test designers are able to handle 11 times more test cases, deliver test results earlier and cover more requirements.

ADAS testing advanced: 6D target mover

Dierk Arp
Executive director
Messring Systembau MSG GmbH
Pedestrians and cyclists account for a significant proportion of road deaths worldwide. Current ADAS test systems are tackling this challenge, but are limited in their design to linear or two-dimensional motion. With this setup, particularly during acceleration processes, an unrealistic motion is generated. The concept of hanging dummies from above creates new possibilities for more life-like dummy trajectories using six degrees of freedom. The system sets new standards in precision and repeatability through the ability to reproduce real-life human motion sequences and imitate them realistically – for example, based on data from a motion capture system.

Autonomous vehicle fleet testing – or how to deal with really big data

Dr Sebastian Bode
Team manager
Leveraging the potential of fleet testing campaigns of autonomous vehicles equipped with multiple sophisticated sensory systems such as lidar as well as communication systems connecting the vehicle with the IoT, challenges even the most bleeding-edge big data technologies. Getting the most valuable insights into your data and optimising your engineering processes will require tools and services for data acquisition, management, analytics and visualisation. This talk will discuss our testing along the way to a solution that combines the best from the world of big data, cloud, AI and automotive engineering.

Room C V2X Testing & Validation
09:00 - 12:45

Time-sensitive networking for autonomous driving

Thomas Schulze
Head of testing, automotive
Spirent Communications GmbH
Time-sensitive networking protocols and applications are the enablers for using real-time functionalities in different industries based on Ethernet/IP networks. For the automotive world, TSN will help to implement driver assistance and autonomous functions in next-generation vehicles. The presentation will illustrate use cases of time-sensitive networking in vehicles and also the ways to validate the network components such as ECUs and end devices like cameras and sensors, as well as the software applications using this technology. Learn how autonomous vehicles act like data centres on the road and how to avoid safety issues based on malfunctions.

Managing risks of autonomous vehicles

Kwang Sheun Tham
Emerging technology lead
IAG Firemark Labs
Autonomous vehicles (AVs) have been pitched as being able to significantly improve safety. However, a mixed fleet scenario with AVs and conventional vehicles is deemed a reality for at least the next 10 to 15 years. This poses safety risks arising from interactions between these vehicles. We use data from trials around the world as well as our own analysis to demonstrate reasonable metrics that can be used to evaluate AV risks and human trust in the AVs.

V2x communication – a test tool’s point of view

Thomas Löffler
Vector Informatik GmbH
V2x communication is a key enabler for autonomous driving and helps extend the functionality of advanced driver assistance systems (ADAS). The presentation introduces the limitations of today’s ADAS systems, and shows why V2x communication is such a hot topic and how it can support ADAS and autonomous driving. It also discusses the measurement and testing challenges coming up with ADAS applications supported by V2x communication.

Testing V2X systems by conducting simulations and field tests simultaneously

Dr Fatih Özel
Project manager and consultant
Oecon Products & Services GmbH
C-ITS systems using V2X communications provide opportunities for achieving sustainable transportation. However, before any V2X system is fully developed and deployed, an extensive phase of testing is required. Using dedicated private test tracks and simulations are two main approaches for testing V2X systems. In this paper, a method is proposed for testing V2X systems by using both approaches concurrently and utilising systems such as stereo cameras and roadside stations. The method is aimed to be implemented on a motorway section of the A39 near Brunswick, which is part of the Lower Saxony Test Field for connected and automated driving.

GNSS jamming and spoofing

Moshe Kaplan
GPS Dome Ltd
During the last few years, the phenomena of jamming and spoofing the GNSS signals has become widespread. The source of jamming and spoofing may be on a national level and on a personal level. GNSS jammers/spoofers are easily obtainable over the net. The future of autonomous driving relies heavily on the availability of GNSS signals, both for navigation purposes and for sensor synchronisation, as well as for V2X synchronisation. It is essential to design autonomous vehicles for operation in jamming/spoofing environments as well as including such facilities in the testing range. GPS Dome provides commercial solutions for jammers/spoofers.

Autonomous drive and connected cars: a VTOX business case

Stéphane Barbier
Chief development officer
The speaker will present a Transpolis business case with an OEM and a traffic-light manufacturer. GLOSA = knowing the position and the current and future status of the next traffic lights to enable the definition of an optimal eco-driving strategy. The driver is advised with an optimal speed in order to get a 'green wave' (pass the traffic light with a green light state), or with a slow-down and stop instruction. The objectives are to reduce stop times, reduce unnecessary acceleration in urban traffic, save fuel and reduce emissions.

Best Practices

Cognitive simulation of the driver for autonomous driving system tests

Jean-Charles Bornard
Cognitive engineering and human factors
ESI Group
ESI is currently developing a cognitive driver model, which will be integrated into its simulation solution Pro-SiVIC, in partnership with the French Laboratory of Ergonomics and Cognitive Sciences applied to Transport (IFSTTAR-LESCOT). Even though human behaviour and cognitive processes are still unsolved questions, human driver cognitive simulation will allow a better understanding of the driver’s state and behaviour. Driver cognition simulation will give autonomous vehicle engineers the ability to test their systems from main concept to HMI specification (human-machine interaction) or system validation, and these tests will integrate key performance indicators of human-like performance and situation awareness.

A case study on accelerating lidar development and capability

Mark McCord
Vice president of engineering
Lidar was used for distance measurement shortly after laser was invented in the 1960s. It remained an exotic metrology technique for almost half a century, until recent development in laser components, together with the intense interest in its autonomous vehicles applications, elevated lidar to the level of 'enabling technology'. In this presentation, we illustrate the complementary capability between radar and lidar while attempting to categorise the latest available lidar techniques by their wavelength, measurement mode and imaging mechanisms.

ADAS development on a driver-in-the-loop simulator

Jelle van Doornik
Product manager ADAS
Cruden BV
Recent research shows that many people are not well informed about ADAS in their cars. This results in unnecessary and unsafe situations as they learn to use the systems while driving, or do not use the available systems at all. Companies focus on sensors and forget about the most important aspect: the interaction between driver and vehicle. Driver-in-the-loop simulators can be used to create a smoother shared control experience and better transitions when reclaiming control. This enables automotive engineers to create better assistance systems, ultimately resulting in an increase in road transport safety.

Could we use DevOps methods to deliver autonomous cars?

Guillaume Belloncle
Smart safe and connected car solution experience manager
Dassault Systèmes
Model-based systems engineering (MBSE) methods have been used for decades by the aerospace industry and are increasingly being adopted within the automotive industry. However, they are sometimes seen as too complex to continuously deliver disruptive innovation for mass markets. Internet companies are continuously delivering innovations using agile software development methods (a.k.a. DevOps), but these methods are limited to delivering safety-critical cyber-physical systems, like a self-driving car. We will outline how mixing these MBSE and DevOps views into an agile MBSE approach can better support the virtual development and validation of autonomous driving technology, leveraging existing industry standardisation initiatives.
Please Note: This conference programme may be subject to change


Autonomous Vehicle International magazine