Toward Machine Metabolism: Design Of A Truss Reconfiguring Robot
Biological metabolism is the process by which an organism breaks down food into its constituent modular elements (catabolism) and then uses those raw materials to create new tissue (anabolism). Metabolic processes demonstrate interesting properties that are difficult to replicate in synthetic structures, such as continual reuse of modular elements in new organisms, autonomous disassembly and assembly processes, self repair, continuous adaptation to functional requirements, and robustness to resource fluctuations. Duplicating these properties in a robotic ecology and composing, decomposing, and then recomposing items out of such modular elements could have a wide range of applications, ranging from infrastructure recovery to space exploration . This is a long-term goal. Many challenges present themselves. These include: the mechanical challenges, such as creating appropriate modular building blocks, designing a reconfigurator robot capable of manipulating these building blocks, and the low-level control and sensing necessary for such a robot; automated design challenges, such as designing solutions based upon functional requirements, creating a flexible and user-friendly functional requirement language, determining the availability of current resources, methods to change and adapt solutions based upon currently available compositions of building blocks, and on-the-fly adaptation to resource fluctuations and assembly errors; and pure algorithmic challenges, such as path planning, efficient decomposing and recomposing, and collaboration amongst a group of reconfigurator robots. This thesis focuses on the questions of mechanical design and begins to explore questions of automated design and path planning. Collaborative and distributive questions will be approached in future work. Within this focus, I have designed and constructed a robotic system to serve as a testbed for exploring these ideas, including: A reconfigurable truss system designed for robotic manipulation. These are the basic building blocks that metabolism acts upon. Five designs are presented, including two variations of the final center-threaded strut element design. Four design iterations of a hinge robot capable of the following: translational and rotational movements along truss elements, and twisting of truss elements. Statistics for several movements on a truss structure are shown in section 3.8. Additionally, I have done work regarding the algorithmic questions of machine metabolism. This includes the creation of an evolutionary algorithm that aimed to evolve a direct instruction set for the reconfiguration of a truss structure into a new structure satisfying given functional constraints. Finally, I was a collaborator on the following related works: With Lobo et al. , a evolutionary algorithm that uses construction trees to represent the reconfiguration process and reconfigures trusses from an initial structure into a final structure, again satisfying functional constraints. With Yun et al. , a reconfiguration planning algorithm that takes as input a source and destination structure and plans an optimal path for the reconfiguration hinge robot to perform the reconfiguration. The combination of these mechanical and algorithmic contributions will yield a demonstrated metabolistic process, and allow further exploration and discovery of the ideas of machine metabolism.
dissertation or thesis