This presentation provides an overview of the ongoing activities of the European Commission in the field of assisted and automated driving. It will explain the regulatory framework, describe the current status and highlight current challenges and future work.

Removing the human driver fundamentally changes what we actually mean by acceptable safety. A simplistic 'safer than human driver' positive risk balance approach must be augmented with additional considerations regarding risk transfer, negligent driving behavior, standards conformance, absence of unreasonable fine-grain risk, ethics and equity concerns. Current standards frameworks and accompanying definitions are likely to be inadequate to ensure safety due to implicit assumptions that are violated when the human driver is removed. A framework relates risk to acceptable safety in a way that is applicable to all autonomous systems.

What the audience will learn

• Why positive risk balance will not give socially acceptable safety
• Examples of safety issues in the news beyond positive risk balance
• How we need to define acceptable safety for at-scale deployments

The hype surrounding autonomous vehicle development has seen a decline in recent years, as the industry has come to realize that mass production of full self-driving technology remains distant. However, productization of autonomous vehicle components, like ADAS Level 2, is advancing steadily within the latest multirange electric vehicles. This presentation brings a business strategy perspective to a tech audience, fostering interdisciplinary discussions on current market entry barriers in productizing autonomous vehicle technology and exploring strategies to accelerate the technology’s commercialization.

What the audience will learn

• Key market entry challenges for autonomous vehicle and ADAS software
• Strategies to overcome regulatory and commercial barriers in AV tech
• Insights on accelerating the commercialization of autonomous technologies
• The importance of interdisciplinary collaboration for future mobility
• Business-driven approaches to integrate AV tech in electric vehicles

The concept of Operational Design Domain is generally accepted to be an essential part of development, testing and approval of Autonomous Driving Systems but still a new concept without real best practice. We show how the idea comes true by contributing to standardizing a human- and machine-readable technical description format, building a proper taxonomy taking requirements of relevant stakeholders into account as well as implementing software to develop and manage complex ODD definitions.

A global team of 52 participants from 21 companies is developing ASAM OpenMATERIAL, a standard for 3D model and material properties to ensure simulation consistency. Releasing in March 2025, it defines physical properties (refraction, roughness) and structural hierarchy, supporting dynamic elements like wheels. Integrated with ASAM’s OpenX ecosystem, it promotes interoperability for movement, positioning and layouts, enabling reusable assets and enhancing sensor-specific simulations (e.g. lidar). The scope includes virtual traffic models but excludes environmental conditions, targeting robust, consistent simulations across automotive applications.

What the audience will learn

• ASAM OpenMATERIAL standardizes material properties and 3D model structures for accurate, consistent simulations
• ASAM OpenMATERIAL integrates with ASAM OpenX standards, supporting interoperability in simulation setups
• Standardized data improves sensor-specific simulations for lidar, radar and perception testing
• Practical applications and scope, focusing on virtual traffic models, excluding environmental factors like weather effects

This presentation explores the indispensable role of periodical technical inspections (PTIs) in maintaining the safety, reliability and compliance of vehicles equipped with advanced driver assistance systems (ADAS) and autonomous technologies. As these systems revolutionize transportation, regular inspections are essential to ensure their continued effectiveness and public trust.

What the audience will learn

• How PTIs ensure safety and reliability in ADAS and autonomous vehicles
• Understand how PTIs ensure compliance with safety and legal standards, reducing risks and building trust in ADAS and autonomous vehicle technologies
• Discover how PTIs prevent system failures, reduce repair costs and enhance performance through proactive maintenance of ADAS and autonomous vehicles
• Learn how PTIs build trust in advanced technologies and drive innovation by supporting data-driven improvements in ADAS and autonomous systems

Many companies in the ADS landscape see the complexity of regulations and standards as a major hurdle to efficiently create value in B2C and B2B contexts. While the ADS systems themselves are becoming more and more intelligent, the safety certification staff’s toolset does not leverage the power of connected information graphs and AI-enabled automation for creation and processing. This talk will introduce the major building blocks required to achieve digital continuous ADS safety certification and will give insights from many years of applied research to explain what the digital ADS safety certification ecosystem of the future could look like. Specifically, the relationship between approval and liability context, risk-based safety cases, the safety engineering work products as well as post-deployment safety monitoring will be analyzed in the context of digitalization needs and potential.

AI-driven systems introduce dynamic learning, adaptability and continuous updates, posing significant challenges to traditional homologation methods. The objective of this paper is to analyze the existing homologation methodologies, such as the New Assessment/Test Methodology (NATM), and examine how various institutions, including UNECE, JRC and SAE, address AI's incorporation into ADS certification. The discussion focuses on identifying gaps in current frameworks, evaluating the harmonization of principles like transparency, robustness and ethical accountability, and proposing a roadmap for future integration. Ultimately, the paper aims to highlight how harmonized approaches can ensure both innovation and safety in AI-enabled ADS.

What the audience will learn

• AI use cases and current trends
• A harmonized roadmap for AI in ADS homologation
• Key challenges in AI homologation
• Current Frameworks and their applications

The synergy between the evolving automotive landscape and advances in cloud computing, virtualization and AI has not only improved current technologies but also sparked transformative shifts. A solution is sought to tackle current challenges, including enabling testing in complex development environments, managing multiple variants/releases and addressing hardware bottlenecks. Elektrobit’s Test Grid platform solution is designed to meet the growing demand for seamless, platform-agnostic scalability in real and virtual test environments. It unifies all test assets, supports homogeneous invocations, enables effortless integration during early development stages and ensures hot-pluggability with existing CI/CD ecosystems.

What the audience will learn

• Adaptation of tooling to support real and virtual environments
• Achieving platform agnosticism
• Seamless integration into CI/CD ecosystem

The journey from groundbreaking research to scalable development in advanced driver assistance systems (ADAS) is a complex yet crucial endeavor. It involves transforming innovative ideas and findings into practical, real-world solutions that enhance road safety and efficiency. Achieving this requires synergistic collaboration between academia, industry stakeholders and regulatory authorities. By fostering a culture of shared knowledge, aligning objectives across sectors and investing in comprehensive testing and validation processes, we can ensure that cutting-edge ADAS technologies are seamlessly integrated into vehicles. This approach not only accelerates the adoption of advanced ADAS features but also sets the stage

What the audience will learn

• Software-defined vehicle layer model
• Current development status in ADAS/AD
• Where research and development need to work together more closely
• What the future of development and research could look like

We will discuss a suite of cutting-edge technologies that demonstrate our commitment to advancing autonomous and connected vehicle ecosystems, specifically tailored for the Indian and global markets. The software-defined autonomous vehicle provides a flexible and scalable foundation for integrating various vehicle functionalities. This software-centric approach allows for continuous SOTA and FOTA upgrades, enabling vehicles to adapt to new technologies and regulations without requiring hardware changes. Autonomous driving stacks are optimized for diverse road conditions and include advanced sensor fusion, real-time data processing and machine learning algorithms that enhance vehicle safety and performance.

What the audience will learn

• Software-defined autonomous vehicle features
• Autonomous driving stack – sensor fusion, perception, path planning and control
• Features update from the cloud

The concept of a software-defined vehicle (SDV) has become a hot topic across the automotive industry as automakers look for ways to address the complex software needs for functions like ADAS and autonomous driving. Many automakers and industry organizations have turned to open source software for SDV development, which has also led to an increase in contributions back to open source projects and the need for Open Source Program Offices to streamline and organize open source activities. Dan Cauchy, Executive Director of AGL, will discuss the current state of SDVs and the work being done by automakers and Tier 1s as part of the AGL SDV Expert Group. He’ll also provide insight into the driving trends behind SDVs and enabling technologies including virtualization, containers, and the cloud. Additionally, he’ll share updates from the recently launched AGL Open Source Program Office (OSPO) Expert Group, led by Toyota to help other automakers set up OSPOs, exchange information, and develop best practices

In the current work we propose a method to apply generative AI GAN methodology on a hardware-aware NN for lidar data. In order to improve the network semantic segmentation performance, we introduce a GAN network into the segmentation. The adversarial training encourages the model to generalize better by simulating a diverse range of conditions, improving its robustness in real-world applications. This increases the performance with regard to classical real and simulated data based training approaches. Once the NN is trained, we also apply several compression techniques to optimally reduce the latency of the NN for real-time processing of dense point clouds.

What the audience will learn

• Lidar
• Embedded NN for dense point clouds
• Generative adversarial networks

The AutoSeg Vision Foundation model is an artificial intelligence framework powering camera-based scene understanding for autonomous driving, developed by the Autoware Foundation. AutoSeg’s custom neural network architecture processes images and computes multiple perception outputs including semantic segmentation, 3D scene estimation, end-to-end path prediction and lane detection – forming the building blocks of autonomous driving. Combining AutoSeg with Autoware’s OpenADKit -offers a modular, scalable and cross-platform AI system which can be trained in the cloud, deployed on the edge and updated over-the-air as part of a holistic data pipeline, enabling autonomous cars to learn as they drive.

What the audience will learn

• Neural network architecture of the AutoSeg Vision Foundation Model
• How AutoSeg performs in real-world driving scenarios
• OpenADKit framework
• AutoSeg and OpenADKit full AI pipeline
• Autoware’s future vision for autonomous driving

Building a fully trainable AD stack is a promising approach to efficiently build and scale an AD system. Transformer neural networks using tokenization of diverse information into a common representation and attention mechanisms to efficiently combine data in time and space foster understandable and flexible end-to-end architectures. An enabler to follow the fast-evolving research in end-to-end is a toolchain allowing quick application of such architectures from training over simulation into closed-loop real-world vehicles. The presentation gives an overview of the toolchain and presents insights gained from tests in simulation and closed-loop applications in the real world.

What the audience will learn

• An overview of the end-to-end state of the art
• Why transformer neural networks have a huge impact on end-to-end
• An example of a toolchain fostering rapid end-to-end experiments
• Challenged and blocking points getting end-to-end architectures into a vehicle
• Insights into first closed-loop experiments with a closed-loop end-to-end AD stack

AI tools are increasingly being used in cyberattacks and defense mechanisms. This is shifting the paradigm of cyber defense by training the LLMs to simulate attack scenarios based on the industry and organization context. This includes the use of business data from the OEMs and Tier 1 suppliers to train the AI modules. Application testers can use generative AI to simulate sophisticated cyber attacks; emulate the exploitation of code configuration issues for preventative security; impersonate the system user by querying mechanisms for human interaction; or simply add context to threat technique identifiers within the Mitre Att&ck framework.

What the audience will learn

• The concept of AI and LLM in security testing
• Use cases
• Kill chain analsysis

From ChatGPT to generative world models for autonomous driving (AD), data for AI model training must be high performance, flexible and scalable. Data helps AI models identify and extract meaningful features from input data. The quality and depth of training data significantly affects the success of AI models. Training data provides examples and relevant information for AI models to learn from. This presentation will discuss AI-powered computing for AV development, data-driven development, and ways to accelerate AV development.

What the audience will learn

• Data for AI model training must be high performance, flexible, and scalable
• Quality and depth of training data significantly impact the success of AI models
• Data helps AI models identify and extract meaningful features from input data

The presentation will focus primarily on the evolution of AD stack from RL algos to diffusion models, and its impacts on robotics and, if data collection is not a challenge, then into autonomous driving.

What the audience will learn

• Google and Alphabet
• ML to AI in ADAS
• Recent advances in computing
• LLMs in motion control
• Outlook

This presentation will explore the role of SOTIF (Safety of the Intended Functionality) in ensuring continuous safety in autonomous trucking. Plus will share its strategy for ensuring the safe deployment of autonomous trucks across different scenarios and environments.

What the audience will learn

• Understanding the importance of SOTIF in autonomous trucking
• How Plus integrates SOTIF principles into its system development
• Key challenges and lessons learned from real-world safety testing

End-to-end AI (e2e AI) is drawing attention in the automotive industry, yet questions remain about its safety and practical application in Level 2+ to Level 3/4 driving systems. In this presentation, hear directly from Wayve, the pioneer of e2e AI, on how it is putting this transformative approach into practice. Learn how Wayve is addressing safety concerns without relying on traditional rules-based methods and why e2e AI excels at solving edge-case scenarios. Discover how this innovative approach is unlocking the path to safer, scalable and adaptable autonomous driving systems.

What the audience will learn

• How Wayve’s e2e AI approach addresses safety
• How Wayve’s approach does not rely on traditional rules-based methods
• Why Wayve’s AI is uniquely capable of solving edge-case scenarios
• Why this is a critical challenge for autonomous driving technology
• How Wayve’s AI is scaling systems ready for real-world deployment

Unlike traditional modularized autonomous driving based on high-definition maps, the end-to-end model DeepRoute IO introduced by DeepRoute.ai doesn't need high-definition maps and has been rapidly adopted by auto makers for series production. It's more human-like, and can handle complex scenarios while ensuring safety. In just four months, over 30,000 vehicles integrated with IO have been released for consumers to enjoy the convenience of smart driving. The presentation will share the paradigm shift brought by the end-to-end model and technology evolution behind it.

What the audience will learn

• Autonomous driving technology evolution in the past decade
• Why end-to-end model is better, what it's capable of
• Commercialization path for end-to-end autonomous driving