In the current world of automated vehicles (AVs), the primary goal is to achieve high safety, while often sacrificing usability. In particular in urban environments, human intervention is often required to maneuver the AV out of a situation. The reasons for the urban challenges are manifold and include high levels of occlusion and a high degree of uncertainty for the behavior of other traffic participants. To address these challenges, a consortium consisting of Intel Deutschland GmbH (Intel Germany), Schaeffler Technologies AG & Co. KG, Karlsruhe Institute of Technology (KIT), FZI Forschungszentrum Informatik, ANavS GmbH, Ibeo Automotive Systems GmbH, und Lake Fusion Technologies GmbH (LFT) started a joined research project called “SafeADArchitect”. SafeADArchitect will be funded by the German Ministry for Economic Affairs and Energy for 2.5 years, with an official start in November 2020, and an expected end in March 2023. The goal of the “SafeADArchitect” project is to develop new approaches and concepts to improve safety of automated vehicles in urban environments. For this purpose, new methods that enable realtime monitoring and mitigation of risks (e.g. collision risk, risk of control loss, etc.) will be in focus. In this regard, it is a primary goal of “SafeADArchitect” to develop a comprehensive architecture solution including necessary safety layers covering the complete system. This means that not only the main software components of the AV are included in this study, but also the compute hardware itself, the necessary sensors and even the driving platform will be considered. In addition, new ways to facilitate the certification process for functional safety standards such as the ISO-26262 will be investigated. The results of “SafeADArchitect” will help to enable and facilitate the development of safety concepts for future AVs.
SafeADArchitect will research an overall system architecture that takes into account uncertainties and risks at various levels in order to safeguard autonomous vehicles. The architecture thereby includes all components from the vehicle platform to the exteroceptive and proprioceptive sensors, the environment perception and the decision making process through to the vehicle's actuators. Therefore these components are extended to enable a continous risk assessment, monitoring and minimization. The risk assessment also allows for the evaluation of the expected risk of different maneuver options and to make an optimized decision after balancing them against an allowable overall risk. As a result, the autonomous vehicle always acts in a risk-sensitive manner and does not make decisions that exceed a given maximum risk. An independent risk assessment makes it easier to approve new components, since their inputs and outputs are examined for risk. This means that systems can be developed to meet specific requirements for certain scenarios without having to carry out a new overall assessment of safety compliance in each case. The concept is depicted in the figure below.
If you are interested in the project or have any questions, please contact:
Dr.-Ing. Fabian Oboril
Intel Labs Germany