Colloquium Cognitive Systems
Toward Knowledge-Based Digital Engineering
Alexander Perzylo, Dr. Markus Rickert (fortiss München)
Abstract. Classical robot programming for industrial robots requires an explicit specification of individual low-level commands in order to achieve a certain result. In this approach, the robot is not aware of the context and overarching goal, while executing its program. Service robots on the other hand are expected to perform complex tasks based only on a high-level instruction. Typically, these instructions are heavily under-specified and formulate a goal without explaining in detail how it can be achieved.
In order to transfer such goal-oriented task specifications to industrial manufacturing processes, it is necessary to formally model various types of knowledge in a way that enables processing by a technical system. Semantic description languages encode common sense knowledge and domain specific knowledge on products, industrial processes, and robot workcells. Based on the product specification and the derived process requirements in combination with the capabilities of the robot system, a solution can be generated automatically.
In this talk, we present our knowledge-based approach to robot programming. It enables intuitive instruction of complex robot systems in a short time frame even for non-expert users. As a result, small lot production, common in small and medium-sized enterprises, becomes financially viable. In an extended digital engineering approach, in which all company data including product and production data is semantically annotated and linked, a fully automated generation of robot programs for a large number of product variations becomes feasible.
Dr. Markus Rickert studied computer science at the Technische Universität München, where he received his doctorate in 2011. He is Head of Robotics at fortiss and has worked on several projects in the field of human-robot-interaction and digital engineering, including the EU FP6 project JAST (Joint-Action Science and Technology), the EU FP7 projects JAMES and SMErobotics and the nationally funded projects SpeedFactory, FarmExpert, and Data Backbone. His research interests include robotics, motion planning, human-robot interaction, cognitive systems, simulation, visualization, and software engineering.
Alexander Perzylo studied computer science at the Technische Universität München. From 2010 to 2013 he worked as a research assistant at the Robotics and Embedded Systems chair of Prof. Knoll. He was involved in the EU FP7 project RoboEarth, which worked toward a World Wide Web for robots. Since July 2013 he works at fortiss, where he was involved in various research and industrial projects, including EU FP7 project SMErobotics, the nationally funded projects SpeedFactory, BaSys 4.0, and Data Backbone. His research interests include robotics, knowledge-based systems, human robot interaction and natural language processing.