This project is to study the design of shape and motion control strategies for nonprehensile robotic devices (e.g., vibratory surfaces) inducing self-assembly of parts. This includes designing sensor-based motions to automatically (1) sort parts, (2) orient parts in fixtures (feeding) and (3) bring two or more parts into a desired relative configuration (assembly). Such systems are characterized by the complex dynamics of repeated frictional impacts. We describe the behavior of the system dynamics in terms of qualitative features from dynamical systems theory, such as stable fixed points and limit sets, chaotic regimes, etc. This abstraction of the descriptive mechanics allows us to use emergent behaviors, the existence of which are robust to uncertainties in the mechanics model, to design sensor-based motion plans to stabilize a desired sorting, fixturing, or assembly.
Success of this work will lead to formal strategies for automated design of industrial parts feeding and assembly systems that are currently designed by trial and error. Our work on sensor design and sensor-based control could find applications to self-assembly in other domains (e.g., on the micro- or nano-scale where the low-level mechanics are quite different). Formal sensor-based manipulation planning approaches, originally developed in robotics, may be applicable to these other domains.
This project will have an impact on undergraduate education by involving undergraduates in the research. The project will also impact graduate education through Northwestern's IGERT on Nonlinear Science and the course ME 449 Robotic Manipulation taught by the PI. Results of the project will be broadly disseminated through project websites, conference talks, and journal papers.
