A geometric approach to task understanding for robotic assembly operations

A task-description method for recognizing a teaching task and applying it in a different environment is presented. The task is described by a sequence of task states defined by areas generated from the segmentation of C-space. The teaching motion instructed from the operator is transferred to a sequence of task states due to the teaching environment. When executing the task, the system observes the task environment, maps the task states to those of the environment, and generates a new task motion. Simulations are presented to show the effectiveness of this method