A representation for spatial constraints in material removal

Abstract

We introduce a model and a representation scheme for solving a general class of robotic spatial task planning problems. We call this class material removal problems. A spatially constrained mobile robot that actively changes its fully-mapped workspace in a predictable manner, removing material from a predefined subspace, is modelled and represented. We describe a composite grammar representation based on structure grammars with which such a robot can be elegantly described. The representation is able to generate a search space of possible robot moves, from which a symbolic search can be used to select one particular plan for the robot. The formalism is thus able to deal with the geometry in a material removal problem in isolation from other knowledge. We demonstrate the utility of the grammar formalism by presenting a pilot implementation of a general spatial task planner/simulator for material removal problems. The implementation, named Mole, shows how the above formalism can be used to describe the workspace and the robot in a given material removal problem. Mole reads this description and creates a simulation of the robot moving in that workspace, automatically generating a reasonable plan for material removal. The main motivation of this work is to represent the spatial constraints on a mobile mining robot as it digs through its changing workspace, but some other application areas are also discussed.