Robots that Duplicate Themselves : Theoretical Principles and Physical Demonstrations

This dissertation primarily focuses on robotic self-replication including a theoretical framework and quantitative measures that can be applied to self-replicating systems. A new descriptive model for physical replicating systems is introduced based on three sets of components: an initial functional system, a set of resources, and a set of external elements. Robotic self-replication is viewed as a process by which an initial functional robot duplicates itself given a set of resource modules in a bounded environment. In order to assess physical self-replicating systems with respect to their structural properties and performance, two quantitative measures are defined. The first is the degree of self-replication which is a combined measure of structural complexity distribution across the modules and the ratio of the robot's complexity to the complexity of the modules. This quantifies an intuitive notion of many simple parts versus a few complex parts. The second is configurational entropy changes resulting from the self-replication process. Configurational entropy is used to measure the amount of uncertainty in the locations of modules. Entropy is also applied for articulated chains by using Fixman's method to compute the mass-metric tensor determinant (MMTD). This dissertation also presents a further extension of Fixman's method to compute