Output details
11 - Computer Science and Informatics
University of Edinburgh
A characterization of the reconfiguration space of self-reconfiguring robotic systems
<22> Originality: Perhaps the first paper, within the reconfigurable robotics domain, to give an account of the connection between the design of motion catalogues and its consequence for planning complexity.
Significance: Beyond the specific domain where this result was obtained, the significance of this paper is in showing how action sets in a decision problem can be characterized in terms of topological properties of the induced configuration space. This has given us a novel line of attack, to take on the difficult problem of representation learning.
Rigour: The paper presents a novel formalisation, theorems based on graph minor theory and detailed experiments based on a spectral graph theoretic criterion.