Embedded Files
Microrobots and biocompatible nanodevices play a crucial role for in vitro applications, requiring precise manipulation and stability in inert solutions such as deionized water. Current methods for controlling these systems often rely on 2D-electrical manipulation, which provides high precision but is constrained to sub-millimeter scales due to limitations in electrode size and working area. Conversely, magnetic manipulation offers larger-scale control but lacks the precision required for microrobotic systems, particularly in tasks requiring fine spatial resolution, stability, and scalability. I developed a high-precision, scalable electrode system that enables manipulation over decimeter scales. This design incorporates an SLA-3D printed resin quadrupole electrode structure with a Ti-Au thin film, integrated with a linear-rail X-Y-Z gantry system to precisely manipulate colloidal Au nanowires under an applied electric field. This setup achieves repeatable large-scale translation of nanoparticles with micron-level precision.
For more information, see "Precision Positioning of Nanowires: Scaling the 'Electric Tweezer'"
To see iterations, see version_2023, version 2021
Brownian motion (no PI control)
Electrokinetic trapping (PI control)
Rotation (PI Control, alternating AC fields)
Alignment (PI Control, switching AC field)
lrm_10x.mp4Long Range Manipulation
Long Range Manipulation
By utilizing PI control with a mechanical XYZ microscopy stage, you can effectively manipulate nanoparticles over vast distances (mm). To accomplish this, you first electrokinetically trap a conductive NP/NW in a central position, move the stage below it, and then navigate around the maze of obstructions (dust, NP attached to the substrate, etc.) to your desired destination. Here, I used an ASI MS-2000 XYZ stage with a resolution of ~22 nm. Furthermore, even with an electrode separation distance of ~1 mm, the NanoPen, with PI control, achieves an aggregate precision (of the trapping) across X and Y of ~57 nm.
The example on the left demonstrates moving a 4.5 μm Au nanowire (180 to 300 nm diameter) approximately 4.5 mm through DI water. In each direction (N, S, E, W), a value in either red (negative) or green (positive) represents the voltage applied across each pair of parallel electrodes in volts (V). The current position relative to the start point is displayed along the black bar at the bottom in μm. The video is sped up ~10x, but an original-speed version can be found here.
Long distance manipulation of Au NW with PI control
npm_analyzer.mp4Parallized nanoparticle tracking software
NPM Analyzer
I wrote a parallized nanowire tracking software using OpenCV with a user-friendly GUI to extract the net velocities of various nanoparticles under applied external electric fields.
Alongside velocity, this software can calculate the rotational velocity, alignment angle, and size of the particles.
The first iteration worked in realtime, but suffered from performance stalls on large quantities of particles. To mitigate this, I wrote a "light" verison that runs a vectorized analysis in steps, first extracting the positions of the particles, then determine the trajectories afterward. Although no longer real-time, the performance increase was 5 - 10x, depending on the video.
5_9 X_original.mp4To address the need for efficient, automated quality control in manufacturing, my senior capstone project with Sandia National Laboratories focused on developing a compact robotic system for visual defect inspection. The primary challenge was to design and build a device within a 6x6x6-inch volume capable of manipulating various small industrial parts, including gears, springs, and circuit boards, with five degrees of freedom for comprehensive visual analysis. My solution integrates a custom-designed, 3D-printed rack-and-pinion gripper with a belt-and-leadscrew-driven gantry system for precise XY translation and a rotating pedestal for Z-axis and rotational control. For defect identification, I implemented a machine learning model using PatchCore, which effectively performs unsupervised anomaly detection by learning the features of non-defective parts. The final prototype successfully automated the handling and inspection process for all specified components cost less than $400.
PyWordle
According to the New York Times, their editors select words daily from the Oxford English Dictionary. Unfortunately, the OED, with over 600,000 words, is behind a paywall, locked only for researchers with license access. To get around this, I generated a list of 2,250,601 possible combinations of 5 letters based on the phonetic and written limitations of English. After cross-referencing with the OED website using an automated Python script, we found 39,513 valid words. Many of these words, however, are archaic, obsolete, and/or outdated. To adjust for this, I used the 2012 Google Ngram dataset, which describes the frequency between the years 1500 and 2012 every word in the Google Books library appears, to calculate a # instances per million words for every valid word. For example, the word "words" appears, on average, 466.08 times for every million English words.
The Wordle rules are simple:
"absent" letters will never appear
"present" letters are present, but in the wrong location
"correct" letters are...correct
Using this as a filter, we can easily determine every possible word for a given game state, and after sorting by the frequency, we can get what word the Editors, who chose the words, most likely selected.
BattleBot
As part of the UT Austin course, ME 366J, I designed, built, and programmed a flipper-style BattleBot. The primary challenge was to design a robust weapon system that could be manufactured using 3D printing while withstanding high-impact combat forces. My solution, "Big Flippa," utilizes a servo-actuated flipper mechanism, which was selected over three alternative concepts through a Pugh chart analysis for its superior speed and versatility. To ensure structural integrity, the flipper arm was iteratively optimized using Finite Element Analysis (FEA) to minimize stress concentrations, and a 2^3 factorial Design of Experiments (DoE) was conducted to validate that PETG with 30% infill was the optimal material configuration. The final prototype was successfully constructed, but an unforeseen current limiter in the RC receiver caused the weapon system to fail under competition loads, providing a critical lesson in electrical system integration and the need for testing under dynamic conditions. A post-competition analysis identified a simple wiring modification to resolve the power delivery issue
annotated-final 366j.pdfPage updated
Google Sites
Report abuse