Homepage » Program » Contributed abstracts » Real-time control » Oral talks » Computationally Efficient Control Strategy for Large (...)
Computationally Efficient Control Strategy for Large Adaptive Optics Systems

Abstract 162

Submitted by Rufus FRAANJE


N. Doelman(1), R. Fraanje(2), P. Massioni(2), M. Verhaegen(2)


(1) TNO Science and Industry; (2) Delft University of Technology.


Current developments towards extremely large telescopes with apertures of 30 to 42 metres put severe requirements on the enclosed Adaptive Optics (AO) systems. The number of actuators in the primary deformable mirror need to be in the range of 10^3 to 10^4. For imaging of exoplanets with Extreme AO the number of actuators in the instrument’s deformable mirror may even be higher. Together with a control bandwidth of several kHz’s, it follows that the computational demands for real-time AO control in extremely large telescopes will be very high. This paper proposes an efficient control strategy to handle the computational complexity of real-time AO control for large telescopes. The strategy is based on a distributed approach, where an array of local controllers replaces the classical centralized controller. Each controller drives a limited number of local mirror actuators based on the information of a limited number of local sensor nodes. The method consists of a sparse internal model controller and a distributed feedforward controller. The distributed feedforward controller consists of N local controllers each driving one actuator. The distributed controller is updated by means of a sparse adaptive gradient algorithm. It is shown that the computational complexity scales proportionally with the number of local controllers N, which is a significant reduction in comparison with the computational complexity proportional to N^2 of centralized controllers. The paper also addresses the stability and performance in terms of residual wavefront error compared to the performance of the classical centralized control approach.

Documents to download