| Summary: | This thesis develops the algorithmic foundations for applying automated planning techniques to program robots to assemble discrete spatial structures. Benefitting from the robot’s capacity for moving, positioning, and holding elements precisely, robotic assembly aims to neutralize the cost and time impact of increasing demand for non-standard, customized designs using programmable robotics and automated process. Programming robots to assemble structures requires us to reason about the construction sequence and the robotic motions. The critical planning challenge is satisfying both stiffness constraints that limit the deformation of the structure and geometric constraints that ensure the robot does not collide with the structure. Current planning approaches either require a significant amount of human intervention or do not scale to the numeric scale and geometric complexity demanded by construction. As we shift from mass production in manufacturing to mass customization in construction, we need versatile planning tools that can adapt to different structural typologies, off-load tedious human programming work, and involve human expertise when relevant.
This thesis addresses this need by proposing a unified algorithmic framework to formulate and solve assembly planning problems. Our investigations are grounded on three broad classes of assembly planning problems: (1) spatial extrusion, (2) pick-and-place assembly, and (3) robotic assembly with multiple tool changes. For each class of assembly problems, we propose scalable, efficient planning algorithms and test them with simulated and real-world case studies. This thesis demonstrates how algorithmic planning can provide us with a much smoother transition between an assembly design and its final execution on the robot. Based on these sound foundations of the "forward-evaluation" of robotic constructability in various contexts, we finally attempt to "close the loop" - deriving a metric to measure constructability and use it to guide the performance-driven exploration of a discrete design catalog.
|
|---|