Developed a shark tracking algorithm allowing two autonomous underwater vehicles (AUVs) equipped with hydrophones to autonomously follow a tagged shark and collect trajectory data with a resolution of 8 seconds and a mean error of 10 meters.

Deployed the AUV off the coast of Long Beach and Costa Rica. Debugged the robot and collected data in a field setting.

Wrote software to communicate with and integrate outputs from multiple sensors including GPSs, hydrophones, DVLs, compasses, and IMUs into combined datasets used for validation.

First author on research paper Multi-AUV Marine Life Tracking via Single Transceiver Payloads. Presented said paper at AAMAS 2023 ARMS workshop.

Prototyped robot localization and mapping algorithms including extended Kalman filter and particle filter localization, and grid mapping. Led new members through implementing these algorithms as well.

Learned about simultaneous localization and mapping (SLAM) algorithms and implemented FastSLAM in C++ and FastSLAM 2.0 in Python.

Created and documented communications interfaces for software subsystems using the Robot Operating System (ROS).

Learned and utilized the Ruckig motion planning library for C++ to generate jerk constrained trajectories, and to filter out abrupt accelerations. This allowed for ±1/16” precision in tests replicating conditions that caused ±1/4” precision degradation in the field.

Implemented and tuned a guided vector control algorithm to reduce overshoot when following arc segment trajectories.

Worked with the controls team of 3 engineers and gave twice weekly updates during standup meetings on my intern project.

Used TensorFlow to fit hidden Markov models to the shark trajectories to identify and explain behavioral patterns in terms of their swimming speed, turning angle, and other covariates such as water temperature.

Created an interactive web-based map using Kepler.gl to animate these trajectories and visualize other aspects of the dataset.