Fraunhofer Institute for Cognitive Systems IKS

Reinforcement Learning (RL) is learning in direct interaction with its surroundings. The blog gives a brief introduction and discusses how the safety of this process could be increased.

To exceed the simple embedded systems of today, advances in CPS require several properties, particularly adaptability, resilience, and safety. ResilientSOA combines these properties e.g. for AGVs in warehouses.

The online seminar »Safety platforms for autonomous systems« introduces concepts to achieve self-awareness of autonomous systems to solve key challenges such as appropriate diagnosis and safety management. It also shows how such concepts could be transferred into practice and provides an outlook into the future of adaptive autonomous systems.

In their new paper, Franziska Schwaiger and her colleagues analyze the miscalibration of object detection models with respect to image location and box scale. Read more about the results of their experiments in »From Black-box to White-box: Examining Confidence Calibration under different Conditions«.

In their new paper, Christian Drabek and his colleagues identify the main challenges of developing Cyber-Physical Systems-of-Systems (CPSoS) based on several industrial use cases and present their novel approach for designing cloud-based safety-critical applications.

Together with Continental, Fraunhofer IKS was able to create a concept for the dynamic distribution of vehicle functions and develop a technical safety concept that describes an implementation of the identified safety requirements.

Artificial intelligence (AI) has to be able to handle uncertainty before we can trust it to deliver in safety-critical use cases, for example, autonomous cars. The Fraunhofer Institute for Cognitive Systems IKS is investigating ways to help AI reason with uncertainty, one being the operational design domain.

Reconciling safety, performance and reliability in cloud-based cognitive systems is no easy task. Especially in safety-critical autonomous systems such as driverless cars, expectations rightly assume that all three requirements are implemented at the highest level. The Fraunhofer IKS has developed an new approach to solve the challenge: flexilience.

Deep Neural Networks have proven their potential. This applies above all to their use in the laboratory. Their performance in real-world applications leaves a lot to be desired. One major problem is the estimation of uncertainty. A closer look at methods that meet this challenge.

To determine a system’s dependability, scientists at Fraunhofer IKS are developing an adaptive, dynamic and flexible safety analysis process that overcomes the limitations of current methods.