Controls Research with the Bridgeman Lab
With the Bridgeman Lab at Duke, I have been working on improving urban infrastructure by developing new ways to utilize drones with a focus on controls.
Project 1: Magnetic Levitation Systems
Spring 2022 – Fall 2022
My first project with the lab was to improve magnetic levitation systems using current from a coil to produce an upwards force on a magnetic ball. Applications of magnetic levitation include high-speed trains which are predominant in urban areas. In an independent study, my goal was to develop a process/algorithm for determining if missing factors in a system are linear or nonlinear. The process consists of linearizing the system (magnetic levitation systems are non-linear in nature) and developing a Bode function based on the equations of motion. Then, an experimental Bode function can be mapped to the original using experimental data to find a missing linear “gain.” This new gain was then implemented into a PID controller, which finds the missing linear factor if stable. If not stable, as was my experimental system, the system most likely has a non-linear factor missing from the equations of motion. This project occurred during my sophomore fall semester and consisted of weekly meetings with my advisor and time in the experimental lab and asynchronous coding work in MATLAB.
This experience has many applications to urban infrastructure that utilizes non-linear control systems, including magnetic levitation, to reach a target point. Other non-linear control systems include air conditioning systems, drones/UAVs, and railways.
Supervisors: Leila Bridgeman, Amy Strong
Project 2: Drone Control Systems
Spring 2023 – Spring 2025
My three-year project at Duke is with the Bridgeman Lab and is focused on developing and testing a feedback control system for UAVs and developing PID and PD controllers to improve their ability to fit into small spaces and withstand harsh weather conditions. My current focus in the research is on system identification techniques for constants with the UAV devices such as drag constants and moments of inertia.
Spring 2023
During the Spring 2023 semester, I began working on this project with my Ph.D. candidate advisor, Ethan LoCicero, and Lelia Jennings. Our main research task was to build a Pixhawk 4 quadcopter from scratch, making adjustments that would later allow us to implement a feedback control system on to the flight controller. I was the sole person on building the drone and worked on setting up a GPS, telemetry, radio transmitter and receiver, and power distribution system for the drone. By the end of the semester, the drone was ready to fly and the controller was ready to be designed.
Whilst building the drone, my team also began working on a controller that worked with a linearized system of the drones’ dynamics. My early work with the lab group consisted of creating the non-linear simulation of the drone dynamics to use for system identification and testing the non-linear system. By the end of the semester, the team was able to finish linearizing the system and found a linear simulation that matched the non-linear at close to equilibrium values.
Fall 2023
In the Fall 2023 semester, I worked more individually on this project with Ethan LoCicero to complete system identification and fine tuning the PID and LQR controller. The drone that we are using has various parameters such as moments of inertia, drag and lift constants, and mass, that all needed to be measured accurately for the controller to identify which inputs will keep the drone on its designated path. Disturbances that I will be investigating include wind and motor jams. In terms of fine tuning a controller, we implemented the PID and LQR controller and testing it in the drone flying area to fine tune the parameters. By the end of the semester, Dr. Bridgeman and I have crafted a straightforward plan to finishing my research and publishing a paper before my senior year based on the success of my work in the fall.
Spring 2024
During the Spring 2024 semester, Ethan and I are working on deriving the small-gain of the drone system to build a controller based on simulation data. Current controller building relies on fine-tuning a controller which is often random and misguided. By using large amounts of simulation data, an estimate for the small gain of the system can be calculated. This semester I worked with the Duke Compute Cluster and learned how to use large computing systems and parallelization to speed up our differential equation simulation algorithms. I have also been working on making the drone more efficient with new devices, including voltage regulators, relays, and degree of freedom sensors.
Supervisors: Leila Bridgeman, Ethan LoCicero
