Translation of task-level instructions to sensing/actuation skills by a robotic assembling agent

This paper describes a robotic assembling agent that translates task-level, assembly operation, instructions into a sequence of executable skills. We adopt the concept of an agent from the field of multi-agent systems and the concept of a task-level instruction from the field of robot programming. A skill is a parameterized, sensing and/or actuation, action that a robot can accomplish repeatably. We present our approach to this translation problem within the larger context of an agent-based robotic assembly cell (AGRA) and in relation to the proof of concept implementation of an actual, robotic assembling agent. In particular, we discuss the translation of a basic peg in hole, task-level instruction that uses a range sensor and a force/torque sensor into a sequence of skills.