DROIDS project

Reliable In-Vehicle Perception and Decision-Making in Complex Environmental Conditions (EVENTS)

Funder: European Union

Partners: I-SENSE Research Group, of the Institute of Communication and Computer Systems (ICCS), Institute of Measurement, Control and Microtechnology at Ulm University (UULM), SEAbility, Tecnalia, Perciv AI, HITACHI France, APTIV, HITACHI UK, STELLANTIS, TU Delft

Funder: Vehicle Certification Agency

The “GB Autonomous Vehicle Approval Scheme” represents a pivotal initiative driven by the Vehicle Certification Agency (VCA) to enhance the safety and security of Automated Driving Systems (ADS). As the landscape of autonomous mobility evolves, ensuring robust certification processes becomes paramount. This project builds upon the groundwork laid by the Department for Transport’s International Vehicle Standards (IVS) Division, specifically focusing on Work Package 3 (WP3): Scenario Generation, Selection, and Coverage.

The rapid evolution of Automated Driving Systems (ADS) demands robust safety assurance. Rooted in assessing the Operational Design Domain (ODD), the proposed methodology offers a generic framework applicable to ALKS and other ADS types. The project focuses on dual aspects: auditing management systems to ensure credible processes and assessing system performance. Additionally, it lays the groundwork for upskilling VCA engineers, enabling effective application of the ODD assurance framework. Beyond the UK, this initiative contributes to international ADS safety standards.

The project involves three work items. The first two focus on Management System Auditing and Performance Assessment. They involve comprehensive research and the creation of training materials for the Vehicle Certification Agency (VCA) engineers. The training will provide foundational knowledge needed before any Type Approval specific training, which would be part of a potential second phase of the project. The third work item will demonstrate how our flexible and scalable Operational Design Domain (ODD) based assurance framework can be applied to various use cases. This will also provide foundational training material on how the framework can be implemented.

A key component of our project is the development of scenarios, which are virtual ‘what if’ situations that the ADS might face. These scenarios are stored in WMG’s online Safety PoolTM Scenario Database, which currently hosts over a quarter million scenarios.

Our team, consists of experts from various fields including ADS verification and validation, project management, scenario generation, scenario simulation, systems and safety engineering, software architecture, and human factors. The project aligns with the high-level milestones requested by the Department for Transport (DfT) and VCA. The outcomes of this project will not only provide great value for VCA but also contribute to the safety assurance of ADS internationally.

i4Driving

Funder: European Union

Partners: Panteia BV, University of Naples Federico II, Aimsun, Delft University of Technology, Swedish National Road and Transport Research Institute, Automotive Technology Centre of Galicia, ZF Friedrichshafen AG, SwissRE, RDW, Technical University of Munich, University of Aschaffenburg, National Research Council, DENSO

The i4Driving project represents a pioneering initiative aimed at revolutionising the safety assessment of automated driving systems (ADS) within Cooperative, Connected, and Automated Mobility (CCAM) solutions. By establishing a robust human road safety baseline for virtual assessments, i4Driving seeks to address the challenges posed by the integration of automated vehicles into heterogeneous traffic environments. At its core, i4Driving embodies two central concepts: the development of a multi-level, modular simulation library integrating existing and novel human driving behaviour models, and the introduction of an innovative methodology to address the significant uncertainty inherent in human behaviours and use case circumstances. This collaborative endeavour draws upon expertise across various disciplines, including traffic engineering, human factors, data science, and computer science.

The i4Driving methodology employs an agnostic approach to modelling, utilising a combination of data mining and advanced cross-disciplinary techniques to guide the development of the simulation library. Through iterative cycles of model development, calibration, and validation, the project aims to strike a balance between model complexity and error, ensuring the plausibility and credibility of the developed models. A key innovation of the i4Driving project lies in augmenting existing human driver models with a 4D cognitive layer, drawing from diverse disciplines such as Human Factors, Road Safety, Traffic Flow Theory, Psychology, Artificial Intelligence, Statistics, Mathematics, and Social Science and Humanities. Sensitivity auditing and techniques for global uncertainty and sensitivity analyses drive the comparison of different modelling assumptions and formulations.

The project's methodology encompasses both inner and outer model development processes, utilising existing datasets and experimental data collected from driving simulator facilities within the consortium. Techniques for standardising experiments and encoding the heterogeneity of driving behaviours further enhance the credibility and applicability of the developed models. Through interdisciplinary collaboration and innovative methodologies, i4Driving aims to accelerate the adoption of CCAM technologies while ensuring the safety and reliability of automated driving systems, paving the way for a safer and more efficient future of transportation.

Scenario, coverage, & Safe AI

Funder: DENSO

This project is contracted by DENSO to undertake multi-phased research activities on 1) STPA-based scenario generation; 2) ODD-based scenario generation; 3) coverage analysis development; and 4) safe AI.

Based on the system architecture, STPA analysis was carried out using domain experts’ knowledge to explore potential hazards within the system. Furthermore, scenarios were also created using the system’s ODD specification and combined with the system’s behaviour competency. Such created scenarios are hosted within the Safety PoolTM scenario database for testing purposes. A multi-pillar-based coverage analysis technique was also developed as part of the project to evaluate the completeness of the scenario-based testing process. In the last phase, the AI safety topic was explored using the perception module of a system and varying detail levels of the input data.

SUNRISE

Funder: European Union

Partners: TNO, Siemens, ERTICO, TOYOTA, VEDECOM, Renault, Continental, Vicomtech, CVC, IDIADA, Chalmers, RISE, bast, IKA, Deepen, Infineon, AVL, Virtual Vehicle, Universita Di Trento, ICCS

SUNRISE stands for Safety Assurance Framework for Connected and Automated Mobility Systems. The main goal of this project is to develop a harmonised and scalable CCAM Safety Assurance Framework that fulfils the needs of different automotive stakeholders. CCAM stands for Cooperative, Connected and Automated Mobility. This safety assurance framework plays a key role in the project. As shown in the figure below, different stakeholders each have their own specific interest in this safety assurance framework. For example, developers of CCAM systems, but also the entities that certify these systems or that perform research in this field, and even citizens, might have their own specific interest in this safety assurance framework.

At a high level, there are three main components of the safety assurance framework: audit, in-service monitoring and reporting (or ISMR), and performance assurance. Performance assurance focuses on ensuring that the system can perform as intended within its defined Operational Design Domain (ODD), which can be visualised as a set of operating conditions within which the system is designed to operate safely. The audit focuses on ensuring the development process, the tools used, and the management system are adequate. The ISMR focuses on capturing and recording any additional information during the system's deployment for consideration in the system’s future design. All components are part of the safety assurance framework; however, in SUNRISE, we focus on performance assurance using a scenario-based approach and incorporate a virtual environment. Please note that we also have the input layer to the safety assurance framework, which includes the requirements containing ODD and behaviour, as well as other external requirements and test objectives.

The performance assurance workflow comprises three main blocks: scenario, execution, and analysis.

1. Scenario Block: This involves creating, formatting, and storing scenarios. Scenarios can be created using knowledge-based (expert insights) or data-based (e.g., UK STATS 19 dataset) approaches. Scenarios are formatted for different stakeholders using common formats like the ASAM OpenSCENARIO/OpenDRIVE and BSI Flex 1889 and stored in a database such as the Safety Pool^TM Scenario database.
2. Execution Block: Concrete scenarios are obtained and ready for execution, combined with test objectives, and allocated to various testing environments. The framework supports virtual, simulated, and real-world testing environments.
3. Analyse Block: Includes coverage analysis and test evaluation. Coverage analysis assesses scenario coverage from multiple perspectives and explores the scenario parameter spaces iteratively to identify failures. Test evaluation checks if individual tests meet criteria like speed limits and collision avoidance. The combined analysis results determine the system's overall safety assurance.