Efficient Motion Planning for Intuitive Task Execution in Modular Manipulation Systems

Abstract

The design of complex systems in robotics requires modern engineering approaches and intuitive task descriptions. This work introduces an instruction paradigm that goes beyond typical programming interfaces and toward a more intuitive interaction style with a robot system. A newly developed motion planning algorithm avoids exhaustive exploration of configuration spaces and actively balances the exploitation of available information in order to achieve significant performance improvements compared to other state-of-the-art sampling-based planners. Accessible descriptions are realized by a combination of task-based control algorithms with multimodal interfaces and cognitive reasoning modules. Abstraction of kinematics, dynamics, geometry, and hardware aspects result in flexible systems with replaceable components. Reengineering between systems is reduced due to the development of an open source framework with common robotics algorithms.