A framework for self -reconfiguration planning for unit -modular robots

Carl A Nelson, Purdue University

Abstract

A novel strategy for self-reconfiguration of chain-type unit-modular robots is introduced. The strategy uses a graph representation for modular robots, allowing for compact processing of the robot's connectivity information. The main contributions of the research are the incorporation of kinematic and loading limitations in the choice of reconfiguration steps and the development of new graph-matching techniques which are useful not only in modular robot reconfiguration planning but in other engineering applications as well. This allows for generation of multiple candidate reconfiguration paths and the automatic elimination of infeasible reconfiguration steps. This thesis begins with a discussion of the state of the art of robotic self-reconfiguration, followed by an overview of the graph theory which underlies the research. Once this foundation is laid, relevant existing methods for robotic reconfiguration are critiqued, and some subproblems preliminary to reconfiguration are discussed. The new reconfiguration algorithm is then outlined and illustrated through a series of examples. Kinematic and force analysis techniques designed for integration with the reconfiguration algorithm are also presented. The focus then returns to the important and broadly applicable combinatoric problem of graph matching, which must be performed preliminary to the actual reconfiguration planning. Finally, the findings of this thesis are summarized, and examples are given demonstrating potential extensions and further applications of the work presented here.

Degree

Ph.D.

Advisors

Cipra, Purdue University.

Subject Area

Mechanical engineering

COinS