Applying for a free student place in doctoral studies

While testing automated driving systems (ADS) has a comparatively long history of more than 20 years of research, much of that research has been limited to semi-automated or partly automated systems where the responsibility for safe behaviour of the ADS is delegated to the safety-driver whenever necessary. In a fully autonomous ADS, however, there is no safety-driver to which control could be delegated and, thus, all decisions must be made by the built-in machine-learning (ML) component(s). Unfortunately, the measures typically used for declaring the quality of a ML-based software component to be “good enough”, are of statistical nature and highly depend on the choice of the training and validation data. Whether these datasets are sufficiently comprehensive such that previously “unseen” situations are correctly handled – at least at the same level a competent human driver would – is difficult to predict. A potential solution to the problem of validating ML-based systems is to separately and independently define what “safe” operation means. The recently defined “Responsibility-Sensitive Safety” (RSS) model could be the key for developing such a solution. The RSS model defines a core set of rules from which sets of independently monitored behavioural requirements for an ADS could be derived. The main goal of the PhD project is to develop and evaluate a method that translates the RSS model into sets of tests which could be applied by development organisations and certification authorities at different stages during the development of an ADS, e.g., during simulations and during on-road testing to ensure that the ADS does not exhibit unsafe behaviour after deployment.

Micro-Mobility is a fast-developing market since it is perceived as a solution for undertaking the last-mile transportation problem. However, major cities worldwide see this solution as a disturbance to the existing traffic and infrastructure, especially sidewalks, creating additional congestion due to its characteristics and being dockless. To tackle these issues and provide a solution to manage the micro-mobility within the urban areas better, there is a need for reliable localization of these lightweight two-wheeled vehicles. Hence, in this Ph.D. thesis work, we are trying to explore the potential behind the better mapping of urban areas, multisensor fusion, and Data-centric AI to increase the accuracy and reliability of positioning and localization of Micro-mobility vehicles in urban areas.