Resumen de KANT: A Tool for Grounding and Knowledge Management

Miguel Ángel González Santamarta, Francisco Javier Rodríguez Lera, Francisco Martín Herrero, Camino Fernández Llamas, Vicente Matellán Olivera

• The roboticist community divides the knowledge representation and decision-making in the symbolic and sub-symbolic fields. Thus, real-world robotics grounded representation requires specific software techniques for landing continuous and discrete state variables in both fields. This research designs and develops an Open Source tool called KANT (Knowledge mAnagemeNT). It will be used to translate sub-symbolic properties and attributes from the robot environment to a symbolic paradigm in the Knowledge Base, in particular to PDDL. Based on Python, KANT includes mechanisms to enhance the process of knowledge manipulation in real-world scenarios through the use of in-memory or non-SQL databases, which would mean a time enhancement of around 30% of total time in middleware such as ROS 2.