Centralized Motion Planning for a Group of Micro Agents Manipulating a Rigid Object

摘要

In the past few years, research on micro-robotic systems has flourished. Technological limits are being constantly broken and the size of micro-robots is being reduced year after year. In these sizes, a number of control problems arise, which are much more important than in the field of 'macro'-robotic systems. Among these are the limited computational, sensing and acting capabilities of micro-robots, the influence of adhesion forces, the large number of simultaneously acting robots etc. In this work, we are studying aspects of the problem of cooperative-multiagent manipulation of a rigid body, constrained by micro-robot limitations. We model the object-agent contact as a 'ball and socket' joint, and the agents as unicycles. The object of our study is to find algorithms for breaking a given object trajectory into agent control inputs. These algorithms should be simple (i.e. computationally tractable), and should scale up well with the number of agents