Project results are expected to contribute to all of the following expected outcomes:

1. Improved validation of CCAM systems enabled by real and synthetic test scenarios, with the widest possible coverage of traffic situations CCAM systems can encounter on European roads.
2. Efficient provision of relevant test scenarios in a permanently updated and therefore dynamic EU wide database.
3. Accelerated AI development and training making use of the dynamic scenario database.
4. Use of the most appropriate approaches (e.g. vehicle-based versus (quasi-)stationary sensor units) to record relevant traffic data, as a basis for the derivation of test scenarios, in different traffic environments according to extending ODDs.
5. Commitment from key stakeholders to the validation methodology, the scenario database and its usage and to the provision of significant volumes of raw data and/or scenarios extracted from such data.

Higher levels of CCAM require validation methodologies making use of scenario-based physical and virtual testing, thereby complementing real-world test drives on public roads, audits and in-use reporting. Scenario-based testing is necessary as conventional testing and validation approaches would require driving hundreds of millions of test kilometres before new CCAM systems or system updates can be deployed. The development of common scenario-based validation methodologies is the subject of HORIZON-CL5-2021-D6-01-02^[1] and should be based on the results of the HEADSTART project^[2]. To enable these common validation methodologies to be widely used, relevant test scenarios need to be provided. These scenarios can partly be defined based on expert knowledge, which, however, needs to be complemented by the extraction of test scenarios from real traffic data^[3], from collision data and in the future, from advanced traffic simulations. The aim of this call topic is to generate a wide range of test scenarios for the training, testing and validation of CCAM systems with a focus on urban and rural traffic, for which there is significantly less knowledge on relevant scenarios than for motorway driving.

To maximise the outcomes, proposed actions should demonstrate upfront commitment from key stakeholders to the validation methodologies, as developed and used in HEADSTART, in a project to be funded under HORIZON-CL5-2021-D6-01-02^[4], in L3Pilot and in Hi-Drive, either by providing significant volumes of raw data or by providing scenarios extracted from such data making use of the automated processing chain. Furthermore, stakeholders should dedicate resources to ensure that the scenarios are developed in a manner that maximises their utility also to other entities and their successful integration in their future (virtual) development and testing processes. Proposed actions are expected to share scenarios in an openly accessible EU wide database, which should be established by a project to be funded under HORIZON-CL5-2021-D6-01-02^[4].

Scenarios and other data shared by stakeholders and existing data made available by national and by other EU-funded projects can be complemented by new data recorded in this action, to provide a realistic set of scenarios with EU-wide coverage.

The proposed actions are expected to address all of the following aspects:

1. AI based tools to transform raw traffic data into reliable, plausibility-proofed data as well as tools for automatic scenario identification and extraction from that data, including the detection of edge cases - the relatively rare, but particularly challenging traffic situations.
2. Generation of variations of scenarios (starting from those based on real traffic data and creating synthetic entries to the scenario database) with a focus on extending ODDs (including adverse weather conditions).
3. Integration of the above in an automatic processing chain with standardised, open interfaces to enable the efficient and seamless use of data from different sources. The processing chain is expected to comply with the FAIR principles, should be agnostic to sensor technologies, data providers and traffic environments, and it should provide for the data management and quality assurance through the whole process.
4. Ensuring reliable merging of scenarios from different data sources (different projects, different vehicles and stationary units, different perspectives etc.).
5. Feeding the resulting scenarios in an openly accessible dynamic scenario database, which can be used for the development, training, virtual testing and type approval validation of CCAM systems, and which should be connected to or integrate existing national databases as far as possible.
6. Quality assurance of the database: Defining approaches and methods to handle uncertainty and the possibility of errors that might propagate in the assessment, including algorithms for their quantification.
7. Demonstration, assessment of the potential and upscaling of (quasi-)stationary sensor units to record high quality big traffic data in various environments, as well as under various environmental conditions and to identify relevant scenarios making use of the processing chain. The focus of recording such data from a “helicopter” perspective - as an alternative to the use of vehicle-based sensors - should be on the provision of suitable data in a cost-efficient way particularly in urban areas. This includes the fusion of data from different sensors. Upscaling requires amongst others the definition of hardware and software requirements for such measuring and recording systems. When recording traffic data in urban areas, proposed action should aim at:
   • high geographic coverage,
   • high seasonal coverage including adverse environmental conditions (e.g. extreme weather conditions) and their synchronized recording and
   • coverage of complex traffic environments including the interaction with other road users (e.g. pedestrians, bicyclists, users of personal mobility devices).
8. Evaluating different approaches to identify relevant scenarios on rural roads based on the developed processing chain and on traffic data to be recorded on various types of rural roads. This includes the fusion of data from different sensors. When recording traffic data on rural roads, roads with low traffic density should be covered in addition to addressing the coverage issues above.
9. Exploring the potential of complementing scenarios extracted from real traffic data with scenarios extracted from validated, highly detailed traffic simulations, including the use of AI to generate edge cases and other adversarial driving conditions in such simulations.
10. Development of a mechanism for the continuous generation of updates of the dynamic scenario database, including an arrangement for the organisational set-up, governance and financial management of the required activities and resources.

The research will require due consideration of cyber security and both personal and non-personal data protection issues, including GDPR. The cyber security of the developed processing chain should be demonstrated for training, virtual testing and validation of CCAM systems.

Proposed actions are expected to develop recommendations for harmonisation and standardisation and to feed into on-going discussions regarding EU type vehicle approval rules as well as in the framework of the UNECE.

In order to achieve the expected outcomes, international cooperation is encouraged in particular with Japan and the United States but also with other relevant strategic partners in third countries.

This topic implements the co-programmed European Partnership on ‘Connected, Cooperative and Automated Mobility’ (CCAM). As such, projects resulting from this topic will be expected to report on results to the European Partnership ‘Connected, Cooperative and Automated Mobility’ (CCAM) in support of the monitoring of its KPIs.

[1] “Common approaches for the safety validation of CCAM systems”

[2] https://www.headstart-project.eu/

[3] Traffic data in this context refers to microscopic traffic data that describes a driving situation, incl. road layout, road users with their dynamic behaviour, other objects and environmental conditions.

[4] Ibid.

[5] Ibid.