Micro-scale 2D chiplet position control - a formal approach to policy design.

We address the problem of 2D position control of micro-objects (chiplets) immersed in fluid. An electric field, shaped by an array of electrodes, is used to transport and position chipletsY using dielectrophoretic forces. We use a lumped, 2D, capacitive based (nonlinear) motion model to first learn a bi-modal control policy. The learning process is based on optimization featuring automatic differentiation. Next we formally prove that this policy can be generalized. The spatial dependency of the capacitances are estimated using detailed electrostatic COMSOL simulations. The results shown in this paper show that the realization of a system that can digitally convert a design directly to a physical placement of micro-chips is feasible.