Keynote Presentations and Conference Opening 09:00 - 12:30
Verification approaches for autonomous vehicles
Amir Shah Senior systems engineer Uber ATG USA
Defining good safety, functional and performance requirements for completely driverless cars has been an industry struggle since the inception of self-driving vehicles. How do manufacturers set real targets and goals for performance and safety instead of chasing 'better' for eternity? A combination of traditional big systems engineering and new methods of verification for machine learning systems will help pave the way for manufacturers to responsibly demonstrate 'good driving'. We will discuss the methods used at Uber ATG and where we hope to see the industry move in the near future as live deployments become reality.
Virtual testing for autonomous vehicle validation and certification
Andras Kemeny Expert leader immersive simulation Renault FRANCE
Autonomous vehicle validation for both engineering design and public acceptance will require the development of comprehensive testing and certification methodologies. In view of the large number of driving miles to check for robust and efficient autonomous driving models, the use of driver-in-the-loop and high-performance computing simulation is mandatory. The new validation methodologies to be introduced for virtual testing are under extensive development today by automotive companies and certification bodies. High-performance virtual environments corresponding to the test scenarios are to be built, as well as new validation tools guaranteeing high-fidelity correlation between field and virtual testing.
Dr Jayar Shankar Programme head, Smart Mobility Solution Singapore Technical Committee for Autonomous Vehicles, Institute for Infocomm Research (A*STAR) SINGAPORE
This session covers a set of provisional national standards that was launched in Jan 2019 to guide the industry in the development and deployment of fully autonomous vehicles (AVs). Known as TR 68, Technical Reference 68 will promote the safe deployment of fully autonomous vehicles in Singapore. As an industry-led effort, comprising representatives from the AV industry, research institutions, institutes of higher learning and government agencies, the group has developed standards covering four key areas of AV deployment: vehicle behaviour, vehicle functional safety, cybersecurity, and data formats. Set as a provisional standard, TR 68 will continue to undergo refinement as AV technology matures and with future feedback from the industry.
10:15 - 10:45
10:45 - 11:45
Panel Discussion - Accelerated Development to 'Level 4-5' AV's
As dozens of companies from Europe and China to Detroit and Silicon Valley look to build and deploy Level 4+ autonomous vehicles (AV's) in the coming years, many are looking to specialized horizontal software suppliers to accelerate their internal development process and time to market. This panel will feature executives from four such software companies, that are bridging the gap between the tech industry and the auto industry.
Qasar Younis Co-founder and CEO Applied Intuition Inc USA
Dr Jason Eichenholz Co-founder and CTO Luminar USA
Autonomous mobility is the disruptive technology of our era, and at its core are optical sensing challenges. Getting better data required to operate safely is the key to a driverless future, and this all hinges on new kind of lidar built for self-driving scenarios. Join OSA and SPIE fellow and co-founder of Luminar Jason Eichenholz in a talk on the requirements for self-driving vehicles and a vision for a major breakthrough in lidar.
Invisible integration of solid-state lidar to make beautiful self-driving cars
Filip Geuens CEO XenomatiX BELGIUM
Lidars are still mostly placed on the outside of vehicles. Mass-production adoption will require more aesthetically pleasing integration. Consumers are expecting higher levels of automation without giving up slick car design. As lidars emit and receive specific lightwaves, elements to cover or hide the sensor typically affect their performance. XenomatiX worked together with Tier 1s on integration of its solid-state lidar. A front-view long-range lidar behind the windshield will be presented. The impact of the windshield will be explained, as well as the benefits and challenges of this integration. Example data will be shown taken in real driving circumstances.
Overcoming the validation challenges for smart video cameras
Davor Kovacec CEO Xylon CROATIA
AI-powered smart automotive video cameras are enabling self-driving vehicles to better perceive their surroundings. Autonomous vehicles will have an estimated 10+ cameras that integrate high-resolution HDR video sensors and vision processors. These systems generate tremendous amounts of heterogeneous data and require innovative new approaches to multi-channel capturing of sensor data for validation. The presentation will give an overview of smart cameras with high-speed serial links, and standard data formats to be used by third-party HIL/SIL platforms. System requirements will vary by OEM, and the design requirements need to support a configurable hardware-based platform for non-invasive capturing of unaltered and accurately time-stamped data.
Bringing the power of radar to autonomous driving
Kobi Marenko Co-founder and CEO Arbe ISRAEL
While OEMs ramp up Level 3 vehicle production, recent AAA testing shows ADAS systems are not ready to handle real-world driving. To achieve L3 and higher autonomy, there is a need for a highly advanced sensor to view roads with ultra-high resolution and a wide field of view – in the hope of resolving ambiguities, achieving low false alarm rates, and coping with mutual interference.This presentation will explain why imaging radar is the only technology that can overcome industry challenges, and discuss the role radar plays in the future of vehicle autonomy and mobility.
15:40 - 16:10
OADF – the cross-domain platform driving standardization in HAD
Philip Hubertus Senior manager, product management Here Technologies GERMANY
Automated driving aims to deliver on two key promises to consumers – mobility and comfort – allowing us to make better use of our time when we drive from A to B. Furthermore, increased safety will result in fewer accidents and traffic-related casualties. A network effect that is driven by the exchange of sensor data, traffic/safety information and map data across the automotive industry players is needed. Learn more about the goals and achievements of the OADF with its members Adasis, NDS, Sensoris and TISA – and how you can participate and add to the network effect to fulfill the promise of automated driving.
Lidar Sensors for Autonomous Vehicles - Testing and Validation Demand
Dr Mircea Gradu Sr. Vice President Quality and Validation Velodyne LiDAR USA
There is significant need for the AV industry to identify lidar requirements and standardize how to address them. The goal is to have lidar products undergo testing and validation based on the standards early in their product lifecycle with the results available to automakers and Tier 1 suppliers.
It is necessary to create specialized tests and validation cases that establish standard ways to determine whether lidar sensors can address the industry demand. To be of value to automakers, all lidar sensors need to be assessed by the same gauge.
How FIR technology will bring Level 5 autonomy to mass market
Yakov Shaharabani CEO AdaSky USA
Far infrared (FIR) technology is experiencing a revolution, entering the automotive market to fulfill OEMs' need for a solution capable of detecting and classifying all living and non-living objects in an AV’s surroundings. This presentation will detail the weakness of radar, cameras and lidar, and how these technologies are unable to independently provide AVs with adequate and reliable detection in every environment condition. Moreover, it will assess the technical advantages of FIR and highlight use cases in which FIR and CMOS cameras can be used together to deliver the sensing capabilities needed to deploy full vehicle autonomy to mass market.
Precise location for AV navigation systems
Anselm Adams Co-founder and CEO Albora UK
Current technologies do not provide fast, high-precision location coordinates in order to be able to navigate a Level 5 AV safely and securely. Although several tech companies are working on the problem, no one has been able to provide a cost-effective solution. Albora is working to provide such solution.
Developing the ultimate perception system for autonomous vehicles
Juergen Ludwig Director of business development Cepton Technologies USA
The automobile industry is trying to determine the ultimate combination of perception systems for safe navigation in autonomous vehicles. The presentation will discuss why the ideal system will rely on lidar sensors for 3D mapping of surroundings, cameras for vision, and radar to determine speed and distance. We will share the pros and cons of each system and how they balance each other out for safe navigation. The session will also cover the challenges of mass production, and the techniques the industry is using for developing lidar sensors that are high resolution, low power, compact in size and low in cost.
Simulation - Validation in the Virtual Domain 09:00 - 12:50
Challenges of high-fidelity sensor simulation in SIL and HIL environments
Holger Krumm Product manager dSPACE GmbH GERMANY
Autonomous vehicles are now on the horizon. Functions for autonomous driving have to be test driven over hundreds of millions of kilometers via software-in-the-loop (SIL) simulation to reduce costs and improve test coverage. For development and validation purposes of algorithms based on sensor data, it is necessary to generate synthetic sensor data for the driving simulation. This work presents a powerful integrated and unified toolchain for SIL and HIL testing of autonomous vehicles. It will address challenges for sensor data generation with GPU-based high-performance rendering by state-of-the-art game engines in combination with raytracing. Future developments will also be outlined.
Leverage hybrid scenario generation for AV certification
Rodolphe Tchalekian EMEA pre-sales engineer for autonomous driving ESI Group GERMANY
Autonomous vehicles should handle complex, unusual and hazardous scenarios without human intervention. Their certification will require a balance between real and virtual testing. To support both, dedicated SW tools should be interoperable, and easy to use and integrate into innovative test and validation processes. ESI is providing a standalone scenario generation solution using a high-level description language based on OpenSCENARIO, together with an environment to simulate the output from multiple physical sensor systems for outdoor scenarios that combine vehicles, obstacles, pedestrians, weather and road conditions. This presentation will highlight results from pilot projects involving hybrid testing and AV certification.
Application of 'over-the-air' radar stimulation for vehicle-in-the-loop tests
Steffen Metzner Technology scout ADAS, simulation and control AVL List GmbH AUSTRIA
The main test and validation development environment of ADAS and AD functionalities is still the actual vehicle. This time-consuming and expensive method lacks reproducibility and demands a large number of staff depending on the complexity of the scenario. The DrivingCube AVL approach aims to bring a ready-to-drive vehicle into a virtual environment. This enables functional testing independent from weather conditions and daylight, and improves reproducibility. This presentation will show the application of radar stimulation for vehicle tests based on the results of research projects. It will also provide an outlook on radar stimulation for future radar systems.
Training perception software with simulated test drives
Nicholas Keel Principal market development manager - autonomous vehicles National Instruments USA
There is no question that simulation is the method by which we achieve the billions of miles of driving needed to verify safe operation of autonomous vehicle software. Although we are making great progress in testing path planning and control algorithms, perception testing is still primarily limited to high-bandwidth data recording and playback. This session introduces simulation tools and techniques that augment data playback with simulation that replicates sensors and environments with enough fidelity to test perception algorithms in the lab.
10:40 - 11:10
Pushing the boundaries – simulation vs. the real world
Dan Atsmon CEO Cognata ISRAEL
The greatest challenge for simulation is finding the matrix that compares it to real life. In this session, Danny will present a mathematical matrix that defines this relationship and will show how a proper simulation can be constructed based on deep learning techniques. He will also supply live examples of the Cognata simulation platform.
Keeping it real: using simulation for AV certification
Kirsty Lloyd-Jukes CEO Latent Logic UK
The presentation will provide an introduction to the OmniCAV project, a groundbreaking, multi-year R&D initiative supported by the UK Government to develop a holistic simulation system that enables 'AVs for all'. This project is addressing the most interesting challenges in AV testing, including: Can you use a simulation to certify an AV is safe to drive on real roads? What about different road environments, like rural roads vs. crowded cities? How can AV developers protect their IP but still get the regulators on board? Project partners are: Latent Logic (lead), Admiral, Aimsun, Arcadis, Arrival, Oxfordshire Council, Ordnance Survey, RACE, WMG and XPI.
Driver-in-the-loop applications with sensor fusion to accelerate ADAS/AD development
Roberto De Vecchi Product development manager VI-grade ITALY
Ram Mirwani Director, global business development, ADAS Konrad Technologies GERMANY
As ADAS/AD functionality gains adoption slowly yet surely, the jury is still out on what type of testing is required and, more importantly, how much testing is required. With governmental and focus groups working together on this topic all over the world, OEMS and Tier 1s continue to develop and test ADAS and AD functionality primarily using drive tests and miles on the road as de facto test methods. The paper will discuss how simulators combined with sensor packages for ADAS/AD can be used together in a lab to robustly verify ADAS/AD functionality as a pre-step to ground truthing.
Real-world 3D road scenario corpus for automated driving simulation
Dr Chang Yuan CEO Foresight AI Inc USA
We present a novel approach for generating realistic and comprehensive road scenarios for driving simulation. The road scenario consists of high-definition road map and 3D trajectories of moving objects (e.g. vehicles, bicycles, pedestrians). Both HD map and motion trajectories are captured by our 3D sensor system from the real world, and processed by computer vision algorithms at high accuracy (5cm level). We are building an extensive scenario corpus of on-road behaviors, including unprotected turns, lane changes, merges/exits and pedestrian traffic. Users can call our web APIs to search and download scenario data in the OpenDrive/OpenScenario format into driving simulation environments.
12:50 - 14:00
Test, Verification & Validation Methodologies 14:00 - 18:15
How to calibrate and test ADAS and AD end-of-line
Thomas Weck Global segment manager AVL List GmbH AUSTRIA
ADAS and AD systems require large-scale testing and validation before vehicles can be released onto the market. While new development processes focus on extensive scenario testing in simulation to cover billions of test miles, vehicle-level testing is important to address integration and confirm the robustness of the sense-plan-act chain. One necessary prerequisite is that all sensors be aligned relative to the vehicle coordinate system. Dürr and AVL have partnered up to consistently transfer testing know-how from development into the end-of-line testbed environment, with the goal of enabling complex sensor calibration and functional testing in manufacturing.
Increasing reliability and efficiency of ADAS data collection campaigns
Alexander Noack Head of automotive electronics b-plus GmbH GERMANY
Testing autonomous systems requires a tremendous number of test vehicles in order to cover the relevant mileage. How can we manage the complexity of these tests? Sensor makers have to ensure that the test vehicles always operate properly, the test systems are configured correctly and the latest sensor software versions are installed. Test drivers sometimes cannot manage the complex test setup with different loggers, data converters, hundreds of cables, etc. An efficient test drive campaign can save a lot of money if it is easy to manage. A central, intelligent management system for test drive campaigns would be desirable.
A regulation-compliant safety framework for the approval of automated vehicles
Dr Houssem Abdellatif Global head autonomous driving and ADAS TÜV Süd GERMANY
Currently there are no general regulations or methods for certification or approval of automated vehicles. However, many use cases or initiatives are struggling to operate automated vehicles on public roads. TÜV Süd has developed a universal framework that copes with the heterogeneous local regulations and provides a unified method for the assessment of automated vehicles. This framework considers local roadworthiness regulations, as well as functional safety and cybersecurity, and can therefore be applied in any region in the world. TÜV Süd has enabled ambitious projects to be successful in operating automated vehicles on public roads, without any safety issues.
Defining and assessing Level 4/5 AV capabilities through physical testing
Niels de Boer Program director, CETRAN Nanyang Technological University SINGAPORE
CETRAN is assessing autonomous vehicles to support the issuance of AV licenses in Singapore. As part of the assessment, not only is safety considered but also the capability to drive in conformance with traffic rules, and with capabilities similar to a human driver and the requirements for the awarding of a driver's license. However, driver's license requirements are written in a way that is hard to interpret using normal engineering standards. The presentation will show how the minimum capabilities required for the assessment are defined, and how tests are defined, executed and assessed to confirm adherence to these requirements.
15:40 - 16:10
Test cases for safety assessment of autonomous vehicles
Erwin de Gelder Research scientist TNO NETHERLANDS
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 is proposed by CETRAN in Singapore that employs virtual and physical tests. For the validity of the assessment, the selection of relevant and realistic test cases is crucial. In this paper, a methodology for deriving test cases is presented. Additionally, many examples are shown because the test cases are currently being used to assess the safety of AVs that are (to be) deployed in Singapore.
Using portable scenarios and coverage metrics to ensure AV safety
Yoav Hollander Founder and CTO Foretellix Ltd ISRAEL
It is now common practice to use scenarios as one of the main ways to ensure AV safety. The presentation will emphasize the need for scenarios to be portable across the various execution platforms used (test tracks, specific simulators, etc.), the various stakeholders (OEMs, subsystem creators, regulators, etc.), various use modes (fully random and deterministic) and so on. It will also discuss how coverage and grading metrics play a crucial role in defining scenarios, evaluating the results and ensuring safety.
Safety assessment of automated driver assistance systems using reliability analysis
Maximilian Rasch PhD and engineer Daimler AG GERMANY
Advanced driver assistance systems (ADAS) are being increasingly focused on by researchers and industries as more and more vehicles are equipped with such technology. Testing and validation are essential to reach this goal and release ADAS. Several papers have stated that the mileage required to prove the probability of failure of the system is impossible to reach in field operational tests. Statistical methods such as Monte Carlo simulation combined with software-in-the-loop (SiL) simulation may help to overcome this limitation. The field of reliability analysis provides algorithms and approaches that can be applied to assess ADAS.
Safe autonomous driving: ASTERO pedestrian target and 6D Target Mover
Dr Igor Doric Executive director Messring Active Safety GmbH GERMANY
This presentation will introduce ASTERO, the latest-generation dynamic pedestrian target. Within the TARGETS research project, which was funded by the Federal Ministry of Economic Affairs and Energy (grant number KF2122308DB3), Messring worked in cooperation with CARISSMA Technische Hochschule Ingolstadt on the development of a novel pedestrian target. Based on previous research results, ASTERO is now focused on realistic human motion while providing the required robustness and usability for AEB tests on proving grounds. The presentation will also include the 6D Target Mover test system, which is currently installed on a proving ground in Germany.
Application of scenario-based testing to automated driving systems
Tim Edwards Senior consultant, CAV technologies Horiba MIRA Ltd UK
New CAV applications must be robust to withstand increasingly complex driving situations and diverse environments. The development of methods for the identification and structuring of these potential test scenarios remains a very active research topic. This paper draws on two UK-based collaborative projects, SAVVY and Human Drive, which address L2 and L4 applications respectively. Specifically, we reflect on the experiences from applying state-of-the-art methods to these projects, and mapping system requirements to comprehensive sets of test parameters and scenarios. A further UK project, VeriCAV, has started this year to further develop this work, including the automation of these processes.
Please Note: This conference programme may be subject to change