Learning-Based Control for Optically Trapped Non-Spherical Particles

The physical momentum of light, while negligible in most situations, becomes significant toparticle dynamics when the particle size is on the micron-scale or smaller. An optical tweezers system harnesses the momentum of light to manipulate these small particles by trapping them near the focal point of a sharply focused laser beam. Since the inception of optical trapping in 1986, it has enabled many scientific discoveries at the small scale. Many optical trapping applications involve trapping spherical objects such as cells or spherical beads attached as handles to biological samples. Thanks to the symmetry of a sphere, models for a spherical particle's dynamics in an optical trap can be easily derived. This has enabled the application of modern, model-based control methods to the trapping of spherical particles resulting in significant improvements in experimental outcomes. The trapping of non-spherical particles, for example rod-shaped bacteria or engineered microstructures, is also an important use case for optical trapping. However, the more complex interactions of a non-spherical particle with the potential well of an optical trap do not lend themselves to closed-form models suitable for real-time control applications. This is a significant hurdle for the dynamic control of optically trapped, non-spherical particles and consequently, existing approaches for manipulating these particles rely on trial-and-error or approximating the particle as spherical. In this work the challenge of applying automatic dynamic control to optically trapped non-spherical particles is overcome through the use of iterative learning (IL) control. This is a model-free approach that seeks to learn the dynamic beam position pattern that achieves a desired manipulation on the particle. As an instantiating example, the out-of-plane rotation of a micron-scale cylinder is taken as the control goal. To develop the IL controller design and verify its performance both dynamic simulations and experiments are performed. The simulation results prove the IL controller design concept viable through the successful learning of a beam pattern that rotates a cylinder. Before experiments are carried out, a purpose-built optical tweezers system is constructed and programmed to implement the IL control design. The system design, programming, and calibration are discussed in detail. Experiments carried out on this system show the success of the IL controller design as it reliably and repeatedly learns a beam pattern that successfully and automatically rotates the cylinder. This unlocks new potential in both the productivity and capability of non-spherical optical trapping applications.

Author dc:creator

• Edlund, Connor

• Iterative Learning Control
• Learning-Based Control
• Optical Trapping
• Optical Tweezers

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/11299/280310

OAI identifier oai:identifier

oai:conservancy.umn.edu:11299/280310