Computational Methods for the Design and Control of Microfabricated Actuator Arrays

Manipulation tasks with actuator arrays fabricated in MEMS (micro electro mechanical system) technology can be modeled by the theory of programmable force elds, which describes the forces generated by the actuators as planar vector elds. Given a task (e.g., unique positioning of a part) and a part geometry, we develop algorithms that automatically generate the corresponding manipulation plans as a sequence of force elds. Programmable force elds are employed as an abstraction barrier between applications requiring array micromanipulation and their implementation with MEMS devices. The theory also provides upper and lower bounds (i.e., complexity, completeness, and impossibility results) for manipulation tasks and force elds. We derive a classiication of vector elds, resulting in design criteria by which eecient manipulation strategies and eeective actuator arrays may be developed. This theory is applicable to a wide variety of actuator arrays. When a part is placed on the array, the programmed vector eld induces a force and moment upon it. Over time, the part may come to rest in a dynamic equilibrium state. By chaining together sequences of vector elds, the equilibrium states of a part in the eld may be cascaded to obtain a desired nal state. The resulting strategies work from any initial connguration (pose) of the part, require no sensing, and exhibit eecient planning algorithms. Massively-parallel, distributed manipulation tasks were performed with micro actuator arrays. Two diierent types of MEMS arrays were built using (1) single-crystal silicon electrostatic actuators, and (2) combined thermobimorph-electrostatic polyimide micro cilia. Force vector elds were implemented by ne-grained SIMD (single instruction, multiple data) control of individual actuators that perform cyclic, gait-like motions. We report on manipulation experiments in which the tasks of parts-translation,-rotation,-orientation and-centering were demonstrated using small integrated circuit (IC) dice. The results of the experiments are consistent with the behavior as predicted by the theory of programmable force elds. 1 Figure 1: Sensorless parts orienting using force vector elds: The part reaches unique orientation after two subsequent squeezes. There exist such orientating strategies for all polygonal parts. See URL