Application of Reinforcement Learning Control to a Nonlinear MEMS Optical Switch

I.O. Bucak (Turkey) and M.A. Zohdy (USA)


Reinforcement learning control, microelectromechanical systems (MEMS), MEMS optical switch, and Non-linear control


MEMS (micro electro mechanical systems) have had many applications in Telecommunications Systems and, especially in Fiber Optics transmission. We consider here the robust control of a nonlinear MEMS optical switch, by means of stochastic real valued reinforcement learning. Specifically, the switch system comprising an electrostatic comb drive, suspension, shuttle mass and blade are precisely positioned despite unmodelled dynamics and disturbances. It is shown via simulations of the optical switch system performance that bumpless transfer behavior is possible under the influence of reinforcement learning control.

Important Links:

Go Back