Aspects of sustainable software design for complex robot platforms in multi-disciplinary research projects on embodied cognition

Abstract

Sophisticated robot systems have become an important part in cognition research. On the one hand, autonomous robots are intended to provide a proof of concept for cognitive models. On the other hand, cognition research becomes a source of inspiration in targeting current limitations in the engineering of robust, flexible and adaptive artifacts. In this work, we discuss aspects of software development and integration for heterogeneous robotic systems in cognition research. As we will argue, one important issue is the combination of different computational paradigms within one robot system, which are rooted in the divergent approaches of engineers and scientists. This discussion lead to the introduction of a software framework aiming to overcome some well known problems of sustainable software development in robotics, but particular important for multi-disciplinary and multi-center cognition research projects. The introduced framework is based on well established standards in software engineering and therefore can be considered for a wide range of cognition research platforms and projects. Further on, we will briefly present a robotic setup where this framework is applied. It consists of a manipulator of 14 DOF (degrees of freedom) and an active vision system of 4 DOF. It is part of research activities aiming to model behavior integration and action-selection mechanisms based in large-scale neural networks.