In this paper, we present some conceptual and experimental results obtained from the integration of a Robotics Cognitive Architecture (RCA) with an embedded Physics simulator. The RCA used, CORTEX, is based on a highly efficient, distributed working memory (WM) called Deep State Representation (DSR). This WM already provides a basic ontology, state persistency, geometric and logical relationships among elements and tools to read, update and reason about its contents. The hypothesis that we want to explore here is that integrating a physics simulator into the architecture facilitates the enacting of a series of additional functionalities that, otherwise, would require extensive coding and debugging. Also, we characterize these functionalities in broad types according to the kind of problem they tackle, including occlusion, model-based perception, self-calibration, scene’s structural stability and human activity interpretation. To show the results of these experiments, we use CoppeliaSim as the embedded simulator, and a Kinova Gen3 robotic arm as the real scenario. The simulator is kept synchronized with the stream of real events and, depending on the current task, several queries are computed, and the results projected to the working memory, where the participating agents can take advantage of them to improve the overall performance.
Initial Results with a Simulation Capable Robotics Cognitive Architecture
Lect. Notes in Networks, Syst.
Iberian Robotics conference ; 2022 ; Zaragoza, Spain November 23, 2022 - November 25, 2022
2022-11-19
12 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Cognitive Taxonomy and Task Gradation in Educational Robotics – Preliminary Results
Springer Verlag | 2022
|Parallel Architecture For Robotics Computation
NTRS | 1990
|PRIMAVERA Project - Initial Simulation Results
British Library Conference Proceedings | 1993
|