Optical Micromanipulations based on Model Predictive Control of Thermoviscous Flows
Abstract
High-precision micromanipulation techniques, including optical tweezers and hydrodynamic trapping, have garnered wide-spread interest. Recent advances in optofluidic multiplexed assembly and microrobotics demonstrate significant progress, particularly by iteratively applying laser-induced, localized flow fields to manipulate microparticles in viscous solutions. However, these approaches still face challenges such as undesired hydrodynamic coupling and instabilities when multiple particles are brought into close proximity. By leveraging an analytical model of thermoviscous flows, this work introduces a stochastic optimization approach that selects flow fields for precise particle arrangement without relying on rule-based heuristics. Through minimizing a comprehensive objective function, the method achieves sub-micrometer alignment accuracy even in a crowded setting, avoiding instabilities driven by undesired coupling or particle collisions. An autonomously emerging "action at a distance" strategy - placing the laser scan path farther from the manipulated particles over time - exploits the $1/r^2$ decay of thermoviscous flow to refine positioning. Overall, objective function-based model predictive control enhances the versatility of automated optofluidic manipulations, opening new avenues in assembly, micromanufacturing, robotics, and life sciences.