New results for closed-loop control of reactive ion etching are presented. Using an algorithm derived from nonlinear estimation theory (extended Kalman filtering), etch rate and etch depth are estimated from dual-wavelength reflectometry data. Using the estimated etch rate, we achieve etch rate stabilization using feedback control by actuating forward power in real-lime. Using the estimated etch depth to trigger an end point, we demonstrate an 83% decrease in final thickness variation over timed etches.