Qualitative physics for robot task planning. I. Grammatical reasoning and commonsense augmentations

Abstract

This paper presents a qualitative reasoning approach to synthesizing robot motion plans from task-level specifications. This method relies on representing the knowledge about robot manipulation as a task grammar augmented with a domain-dependent task analyzer. The task analyzer tests constraints for deriving suboperations pertaining to a higher-level operation and performs commonsense reasoning about the manipulator-level motion strategies.A case study on planning and analyzing motion strategies for mechanism-oriented manipulation tasks is presented in the companion paper [ll]. In this paper, through several examples, we show how syntactical knowledge about robot operations can be formulated for real-world manipulation tasks and how commonsense knowledge can be augmented to constrain the syntactical reasoning.