Motion Planning for a Tethered Mobile Robot

Abstract

Recently there has been surge of research in motion planning for tethered robots. In this problem a planar robot is connected via a cable of limited length to a fixed point in R2. The configuration space in this problem is more complicated than the one of a classic motion planning problem as existence of the cable causes additional constraints on the motion of the robot. In this thesis we are interested in finding a concise representation of the configuration space that results in a straightforward planning algorithm. To achieve such a representation we observe that configuration space manifold has a discrete structure that conveniently can be separated from its continuous aspect when it is represented as an atlas of charts. We provide a method for generating either the complete atlas or a subset of its charts based on special cable events. Generating parts of the configuration space on-the-fly enables the following improvements over the state-of-the-art. a) We decompose the environment into cells as needed rather than an off-line global discretization, obtaining competitive time and space complexity for our planner. b) We are able to exploit topological structure to represent robot-cable configurations concisely leading us towards solutions to the more complex problems of interest.

To underscore the potential of this representation, we take further steps to generalize it to two more complicated instances of the tethered robot planning problem that has been widely disregarded in the literature. We will first consider a simplified model of cable-to-cable contacts, giving the robot the option to perform knot-like tying motions. Next, we will address the planning problem for a tethered robot whose cable has a constraint on its curvature. This adds to the realism of the model since most practical cables have some degree of stiffness which limits curvature. In this case we provide a novel technique to relate Dubins' theory of curves with work on planning with topological constraints. Our results show the efficiency of the method and indicate further promise for procedures that represent manifolds via an amalgamation of implicit discrete topological structure and explicit Euclidean cells.