Project Overview

Assuring autonomy is currently a significant barrier for the adoption and the market success of autonomous mobility. Overcoming this challenge is consequently a major challenge and there is unfortunately no simple solution to it. Assuring autonomy based on established methods and standards is either not possible or it implies too many constraints (overly restrictive performance due to worst case assumptions) or infrastructure investments (e.g. a complete overhaul of the traffic infrastructure to address the limitations of today’s automated systems). New solutions are thus required and we see LOPAAS as a key part of such a solution. The consortium has the potential to engineer and assure this solution based on its previous work and its complementary competencies.

LOPAAS could not only be an enabler for autonomous systems in mobility but also in other domains. Autonomous vehicles and other systems have enormous potential to contribute decisively to the solution of the current ecological, social, and economic challenges. They can often perform tasks more precisely, faster, or otherwise better, resulting in ecological benefits. „Precision farming” has becomeindispensable in the agricultural sector; for example, a lot of fertilizer can be saved through precise fertilization. Autonomous systems in mobility could have a huge positive social impact. For instance, it would have a huge impact for the section of the population that cannot drive a vehicle due to age or any other reason. Autonomous systems can counteract the shortage of skilled workers in many areas such as construction, freight and passenger transport, medicine, and nursing care, despite the demographic change. However, as long as autonomous systems cannot adequately manage risks, it is not possible to tap the full potential of autonomous systems. DRM exactly addresses this demand but is challenging to come up with a suitable DRM approach that will be adopted by industry.

The first challenge is to research a DRM approach. Previous work of the consortium provides essential pieces of the puzzle but it is challenging to integrate them into an overall solution. For instance, previous work provides an approach for determining at runtime the uncertainty of ML-based information. Further, it provides methods to manage risks at runtime. It is, however, an open research question how DRM can benefit from uncertain information. DRM is obviously safety-critical. Using uncertain information for realizing DRM is thus a challenging engineering task. Uncertainties need to be minimized and handled so that safe and performant behavior can be achieved. The approach developed by this project will explicitly include uncertainties and their propagation through the system as an intrinsic part of the underlying model.

The second challenge is to transfer the DRM approach to industry. It is very unlikely that industry will adopt DRM without an appealing transfer route starting from the situation where industry is currently in. Industry is confronted with inherited burden concerning safety standards and system architectures. DRM has to harmonize with these inherited burdens. It has to harmonize with the fragmented world of existing safety standards and ongoing standardization activities. This includes standards addressing safety assurance like UL 4600 as well as standards that address also reference safety architectures like ISO/PRF TR 4804.