CAD Webinar Series (III): AdaptIVE Evaluation Framework
On Monday afternoon, Christian Roesener of the RWTH Aachen’s IKA delivered the third installment of the CAD Webinar Series. In his presentation, he covered the methodological foundations of the recently completed European project AdaptIVE, as well as some of its findings. At the beginning, he set out the importance of dynamic evaluation models for connected and automated driving: The evaluation needs to keep track with a rapidly developing set of functionalities, otherwise they will “become the bottlenecks of AD introduction”. Based on this observation, AdaptIVE evaluated not only the system and its functionalities from technical and user-centric perspectives, but also explored the impact of autonomous driving on safety and the environment.
For the technical evaluation, IKA made use of a partially automated machine-learning process to cluster similar driving situations that arose from the large-scale trials. These were then compared against the baseline of human driving behaviour, as collected in previous EU-funded projects. User evaluation focused on behavioural observation and questionnaires, whereas the impact assessment was based on simulations, which extrapolated from the collected data. Over the course of its duration, the AdaptIVE project found for example that the duration of lane changes is much more uniform with automated driving functions, that user behaviour improves when guided by these functionalities and that necessary parking space can be reduced.
At the end of the webinar, Christian Roesener answered questions from the audience, directed the attention to the AdaptIVE public deliverables and thanked the contributors to the evaluation effort of the project. The recording of the third CAD Webinar will be made available in short time. Already next week, the next CAD Webinar will take place. Registration is still open and as always free of charge.