No abstract

In 2016 a collection of guiding principles for the management of scientific data was proposed by a consortium of scientists and organizations under the acronym FAIR (Findability, Accessibility, Interoperability, Reusability). As many other disciplines, control theory also is affected by the (mostly unintended) disregard of these principles and to some degree also suffers from a reproducibility crisis. The specific situation for that discipline, however, is more related to software, than to classical numerical data. In particular, since computational methods like simulation, numeric approximation or computer algebra play an important role, the reproducibility of results relies on implementation details, which are typically out of scope for written papers.While some publications do reference the source code of the respective software, this is by far not standard in industry and academia. Additionally, having access to the source code does not imply reproducibility due to dependency issues w. r. t. hardware and software components. This paper proposes a tool based approach consisting of four components to mitigate the problem: a) an open repository with a suitable data structure to publish formal problem specifications and problem solutions (each represented as source code) along with descriptive metadata, b) a web service that automatically checks the solution methods against the problem specifications and auxiliary software for local testing, c) a computational ontology which allows for semantic tagging and sophisticated querying the entities in the repo and d) a peer-oriented process scheme to organize both the contribution process to that repository and formal quality assurance.

Physical interaction of humans can be a challenging. For some practically relevant situations a reasonable model is given by a closed kinematic chain of a planar rigid body mechanism. From a control perspective a cooperative motion, e. g. assisted standing-up, can be seen as an optimal control problem (OCP) with input and state restrictions. Due to the algebraic constraints and the comparably high number of joints involved, proper formulation of such a problem is not trivial. In this contribution we compare two different modeling approaches: discretization of the full dynamical problem and iterative solution of consecutive stationary problems. The latter turns out to be significantly faster while still providing-in some sense-optimal solutions.

ZusammenfassungIn diesem Beitrag nutzen wir Künstliche Neuronale Netze (KNN) zur Approximation der Dynamik nichtlinearer (mechanischer) Systeme. Diese iterativ approximierten neuronalen Systemmodelle werden in einer Offline-Trajektorienplanung verwendet, um eine optimale Rückführung zu bestimmen, welche auf das reale System angewandt wird. Dieser Ansatz des modellbasierten bestärkenden Lernens (engl. model-based reinforcement learning (RL)) wird am Aufschwingen des Einfachwagenpendels zunächst simulativ evaluiert und zeigt gegenüber modellfreien RL-Ansätzen eine signifikante Verbesserung der Dateneffizienz. Weiterhin zeigen wir Experimentalergebnisse an einem Versuchsstand, wobei der vorgestellte Algorithmus innerhalb weniger Versuche in der Lage ist, eine für das System optimale Rückführung hinreichend gut zu approximieren.