A software framework for the implementation of Dynamic Neural Field control architectures for human-robot interaction

Abstract

Useful and efficient human-robot interaction in joint tasks requires the design of a cognitive control architecture that endows robots with crucial cognitive and social capabilities such as intention recognition and complementary action selection. Herein, we present a software framework that eases the design and implementation of Dynamic Neural Field (DNF) cognitive architectures for human-robot joint tasks. We provide a graphical user interface to draw instances of the robot's control architecture. In addition, it allows to simulate, inspect and parametrize them in real-time. The framework eases parameter tuning by allowing changes on-the-fly and by connecting the cognitive architecture with simulated or real robots. Using the case study of an anthropomorphic robot providing assistance to a disabled person during a meal scenario, we illustrate the applicability of the framework.