Parameters Optimization of the Beam Cleanup System Based on
Stochastic Parallel Gradient Descent Method
Wang Sanhong
*a,b
, Cui Junfeng
b
, Ma Haotong
c
, Liang Yonghui
c
, Yu Qifeng
a
a
College of Aerospace and Materials Engineering, National University of Defense Technology,
Changsha, Hunan, China 410073;
b
Taiyuan Satellite Launch Center, Taiyuan, Shanxi, China 030027;
c
College of Opto-Electronic Science and Engineering, National University of Defense Technology,
Changsha, Hunan, China 410073
ABSTRACT
In a high-energy laser, the thermal aberrations degrade the beam quality and reduce the laser’s output power. Adaptive
optics (AO) technique based on a stochastic parallel gradient (SPGD) algorithm can be used to compensate for the
distortions in real time to clean up the laser beam. Such a beam clean-up system was simulated and experiments were
conducted to study the optimization of the parameters of the gain coefficient and the amplitude of the perturbation. The
results show that the convergence property of the SPGD algorithm is improved after the parameters being optimized.
Keywords: high-energy laser, beam cleanup, adaptive optics, stochastic parallel gradient descent, parameters
optimization.
1. INTRODUCTION
For a high-energy laser, there are many factors causing its beam quality to be lower than the ideal state, such as the
imperfection of the laser cavity, optical cavity misalignment, figure error from mirrors’ manufacturing tolerances, and
distortion from heating of the media on the propagating path. All of these factors cause the wavefront distortion in the
laser’s output beam.
1
These aberrations prevent the laser beam energy from propagating to a longer place or focusing on
a smaller area. So many attempts such as introducing negative lens or optical phase conjugate mirror have been taken to
correct for the aberrations.
2
However, these conventional methods could not completely deal with the aberrations,
especially those aberrations varying with time. As a special technology to correct for the dynamic wavefront distortion in
real time, adaptive optics method is a good choice to clean up the high-energy laser’s output beam.
For the extracavity beam cleanup, which is referred to in this paper, the wavefront corrector of the adaptive optical
system is placed outside the laser resonator. The familiar adaptive optics based on wavefront sensing was once used to
study beam cleanup for an annular beam.
3
However, such systems are bulky and often limited by the phase’s branch
points and the amplitude scintillation which exist in the high-energy laser beam. Another kind of adaptive optics based
on direct optimization of a scalar quality metric was also adopted to clean up a laser beam. The often used optimization
algorithms include generic algorithm, simulated annealing algorithm and stochastic parallel gradient descent (SPGD)
algorithm. And the adaptive optics based on SPGD algorithm is the more promising beam cleanup method because of its
faster convergence. Beam cleanup experiments were conducted by use of adaptive optics based on SPGD algorithm with
iterative rate of 100 Hz and the result showed its success.
4
However, the convergence performance of this kind of system
is affected by two parameters — perturbation amplitude and gain coefficient, which are existed in SPGD algorithm’s
updating formula. Several rules and methods have been given to optimize the two parameters.
5-8
But their values need to
be matched to the unknown wavefront distortion. Because the wavefront distortion from the high-energy laser resonator
is different from that induced by atmospheric turbulence, those rules and methods of optimizing the two parameters,
which are suitable to atmospheric turbulence case, maybe no longer are useful. In the field experiments of beam cleanup,
how to choose easily and shortly the two parameters’ best values is still a very difficult mission.
*
sanhongw@gmail.com; phone 86 351 6015221
High-Power Lasers and Applications VI, edited by Upendra N. Singh, Dianyuan Fan,
Jianquan Yao, Robert F. Walter, Proc. of SPIE Vol. 8551, 855113 · © 2012 SPIE
Proc. of SPIE Vol. 8551 855113-1