Preventive sensing planning for robotic assembly

Abstract

This paper presents our approach of systematically generating preventive sensing plans from the specification of robotic assembly tasks. The sensing strategies are based on the identification of the "critical dimensions" for the assembly tasks. The recognition of such dimensions comes from a model-based task-oriented analysis of the contact relations among the manipulated objects and the environment, and the change in their degrees of motion freedom. The sensing strategies are evaluated by their ability to maximize a predicted success probability.