A wide variety of micromechanical structures (devices typically in the
m range) have been built recently by using processing techniques
known from VLSI industry. Various microsensors and microactuators
have been shown to perform successfully. For example, a single-chip air-bag
sensor is commercially available; video
projections using an integrated, monolithic mirror array have been
demonstrated recently. More difficult is the
fabrication of devices that can interact and actively change their
environment. Problems arise from (i) unknown material properties and
the lack of adequate models for mechanisms at very small scales, (ii)
the limited range of motion and force that can be generated with
microactuators, (iii) the lack of sufficient sensor information with
regard to manipulation tasks, and (iv) design limitations and geometric
tolerances due to the fabrication process.
The design, analysis, and control of micro-electromechanical systems (MEMS) are inherently geometric in nature. While much is known about building MEMS, little is understood about how to use them to manipulate objects. More precisely, manipulation of solid objects subject to hard or intermittent contact is entirely open. The state of the art is somewhat similar to that of robotics before geometric planning algorithms. Development of control strategies for manipulation by MEMS is a key bottleneck. MEMS actuators are tiny; hence one is forced to consider parallel manipulation strategies by teams or arrays consisting of a large aggregate of micro-actuators. It is believed that based on work on sensorless and near-sensorless manipulation, one can develop geometric theories of manipulation and control for microactuator arrays. Can one develop sensorless (open-loop) manipulation algorithms for arrays of MEM actuators? Can these strategies be time-invariant, or is a clock necessary? Can MEMS implement geometric filters? Can MEMS arrays orient or pose parts uniquely, without sensing? Can MEMS arrays implement assembly algorithms?