Automation of alignment tasks can provide improved efficiency and greatlyincrease the flexibility of an optical system. Current optical systems withautomated alignment capabilities are typically designed to include a dedicatedwavefront sensor. Here, we demonstrate a self-aligning method for areconfigurable system using only focal plane images. We define a two lensoptical system with eight degrees of freedom. Images are simulated givenmisalignment parameters using ZEMAX software. We perform a principal componentanalysis (PCA) on the simulated dataset to obtain Karhunen-Lo\`eve (KL) modes,which form the basis set whose weights are the system measurements. A modelfunction which maps the state to the measurement is learned using nonlinearleast squares fitting and serves as the measurement function for the nonlinearestimator (Extended and Unscented Kalman filters) used to calculate controlinputs to align the system. We present and discuss both simulated andexperimental results of the full system in operation.
展开▼