Stochastic Optical Trapping and Manipulation of a Micro Object With Neural-Network Adaptation

Abstract

Optical tweezers are capable of manipulating micro/nano-objects without any physical contact and, therefore, are widely used in biomedical engineering and bio-logical science. While much progress has been achieved in automated optical manipulation of micro objects, the Brownian motion is commonly ignored in the stability analysis in order to simplify the control problem. However, random Brownian perturbations exist in the micromanipulation problem and, therefore, may result in failure of optical trapping due to the escape of the micro object from the trap. In addition, it is usually assumed in the development of a controller that the model of trapping stiffness is known, but the model is difficult to obtain because of its spatially varying feature around the center of laser beam, and variations with laser power and dimensions of objects. In this paper, a neural-network control method is proposed for optical trapping and manipulation of the micro object in the presence of stochastic perturbations and unknown trapping stiffness. The unknown trapping stiffness and dynamic parameters of micro objects, which vary with different laser power settings and sizes of the objects, are approximated by using adaptive neural networks. The stability analysis is carried out from stochastic perspectives by considering the effect of the Brownian motion in the dynamic model. Both experimental results and simulation results are presented.