Phasing segmented telescopes via deep learning methods: application to a deployable CubeSat.

Area of Science:

• Optical engineering
• Space technology
• Artificial intelligence

Background:

• High-resolution Earth imaging from Low Earth Orbit (LEO) typically requires large, costly telescope apertures.
• Deployable CubeSat telescopes offer a compact solution but necessitate precise mirror phasing.
• Limited volume and power on small platforms restrict traditional phasing methods.

Purpose of the Study:

• To develop a computationally efficient method for measuring co-phasing errors in deployable telescopes.
• To enable diffraction-limited imaging from small, cost-effective satellite platforms.
• To overcome the constraints of traditional phasing techniques on compact systems.

Main Methods:

• Development of a neural network (NN)-based algorithm for co-phasing error detection.
• Utilizing a point source for measuring wavefront errors (WFE).
• Testing the NN model's robustness against high-order aberrations and noise.

Main Results:

• The NN method accurately detects phasing errors, achieving target performance levels (WFE < 15 nm RMS).
• The technique demonstrates robustness in the presence of aberrations and noise.
• Performance was validated against existing state-of-the-art phasing methods.

Conclusions:

• A feasible NN-based solution for phasing deployable CubeSat telescopes has been developed.
• This method provides a realistic pathway to achieving diffraction-limited images from small satellite platforms.
• Enables cost-effective, high-resolution Earth observation capabilities.