This award funds an initial investigation of design, control, and planning issues for self-reconfigurable robots. A self- reconfigurable robot is a set of one or more autonomous modules with the capability to self-organize, that is, to change their geometric shape autonomously. Self-reconfigurable robots can be viewed as a minimalist approach to designing versatile, extensible, and reliable robots. They are suitable for applications that require multiple locomotion gaits over different types of terrain, or that combine mobility with flexible manipulation in ways that could not be anticipated at the time of initial design of a conventional robot. The focus of this work is task-directed self-reconfiguration, with an emphasis on robots that can self-organize to support multiple locomotion gaits and on robots that can sequence locomotion and manipulation gaits. This research will explore several designs for self- reconfigurable modules. The potential of each design will be evaluated by geometrically characterizing the class of three- dimensional robotic structures that can be built with it. This research will also consider control and planning approaches to self-reconfiguration. For each module design, control and coordination algorithms will be developed and evaluated. A set of basic configuration structures that consist of the basic module designs and support specific manipulation and locomotion gaits will be developed. Planning algorithms for transitioning between any two such configurations will be developed and evaluated.