Autonomous structure assembly using potential field functions

Abstract

The work in this paper aims to introduce a new technique for artificial potential field methods based on a superquadric representation of the obstacle shapes, and a quaternion representation of the object orientation. The superquadric functions facilitate the representation of the actual shape of the obstacles to provide additional free space for the motion of the maneuvering objects, consequently reducing maneuvering effort for these objects. The quaternion representation overcomes the singularities produced when using Euler angles and is more convenient for real time implementation. Potential field methods provide a robust means of generating distributed controls for mobile robots (terrestrial or space) to enable automated assembly tasks. The problem of local minimum formation when dealing with objects having straight edges is also addressed in this paper through changing the superquadric function shape simultaneously in accordance with the relative position and orientation of the objects. The proposed potential function enables maneuvering objects to decide which motion, both in translation and rotation, leads to the quickest descent of the artificial potential most effectively. These coupled translation/rotation manoeuvres show significant benefits over de-coupled translation maneuvers.