The Governing Council of the University of Toronto

Purpose

The Project involves the application of reinforcement learning (RL) techniques for wavefront sensorless adaptive optics systems used in optical satellite communications. A predictive model will be developed to maximize the coupling of a free-space optical beam into a single mode fiber through an array of wavefront correction elements. Unlike typical adaptive optics for astronomy, optical satellite-to-ground communications experience a wide variety of turbulence conditions and can greatly benefit from capturing a coherent optical signal for improved optical channel data throughput, lowered reduce bit-error rates, and reduced emitter laser power requirements. Simulated wavefront distortion will be used to train the model for optimal performance under the dynamic conditions during overhead satellite passes with full wavefront measurement. A trained RL model will provide optimal parameters with existing AO systems and will guide the design of future photonic AO systems for lower cost and improved performance in a variety of ground station locations.