UC undergraduate collaborates with Swiss university on drone research

The opportunity with ZHAW Zurich University of Applied Sciences in Switzerland came from a college-wide email from former Associate Dean Eugene Rutz who was responsible for facilitating the collaboration. She was not sure if or how she’d make it work with her already busy schedule.

But soon she was on her way to Switzerland, where she connected with Manuel Arias Chao, PhD, assistant professor at ZHAW and Tu Delft.

She worked closely with him and Chetan Kulkarni, PhD, on a project with NASA Ames Research Center, focusing on using audio data to determine if drone motors are faulty. Using a prerecorded set of audio files, Chawla, Chao, and Kulkarni from NASA used machine learning models to predict failures.

“What makes this project different than others is that audio is not commonly being used as the factor to determine fault, but we also are training our model only on healthy data,” she said. “So, anything that does not look healthy is considered faulty.”

In industry, very few drones will be faulty, so there will be a lack of faulty data with which to train drones. If the drones are trained on healthy data, they’ll classify everything as healthy because that’s what the model is expecting. Faults will be missed, which cannot happen in mission critical situations in which the drones will be used. During her semester in Switzerland, she helped develop the model and had it working on the computer.

When she returned to UC, she wanted to continue working on the project as she’d found a great passion for the work. However, for it to be considered an individual study experience at UC, she must be supervised by a CEAS faculty member. Given the project’s area of work, she was provided a list of recommended UC professors whose work aligned. She reached out to Manish Kumar, PhD, professor of mechanical engineering, and even went to his office between classes to speak with him.

“I chose a mechanical engineering faculty member instead of computer science because I already had a CS mentor, and I wanted someone who could help me with a drone and how it works, not just in theory, but actually building it,” Chawla said.

Currently, she is building upon this experience working under Kumar's guidance for her cross-major senior design project along with her team Siddharth Urankar and Ally Blair, where she is recording data about drone batteries, predicting remaining useful life, training models, and putting it all together by building the drone. She plans to present the drone at the annual CEAS EXPO.