Spring 2026
Advisor: Nadia Figueroa, PhD
Co-developed a dual-arm manipulation system in Isaac Sim, designing a joint-space neural dynamical system with Lyapunov-based stability guarantees to control reach-and-transport motions for two Franka Panda arms sharing a workspace. Built a state-dependent modulation layer that treats each arm's end-effector as a moving obstacle for the other, resolving collision conflicts in the shared stacking region without a discrete coordinator making real-time decisions. The system reliably stacked all 6 blocks in 14.4 seconds; in an ablation with the modulation layer disabled, it failed to place a single block before timing out.
Reinforcement learning drone racing
Fall 2025
Advisor: Antonio Loquercio, PhD
Built a reinforcement learning policy that flies a quadcopter through a multi-gate race course using PPO, engineering the reward structure, observation space, and episode reset strategy that determined success. Diagnosed and abandoned a phased-training approach that looked promising in theory but caused policy divergence in practice, then converged on a single-stage training run with domain randomization for robustness. The resulting policy completed full race laps in simulation at a 95% success rate across 20 evaluation episodes with zero crashes, at a mean 3-lap time of 19.9 seconds, and we deployed the trained policy onto a physical Crazyflie drone.
Aug 2023 – May 2024
Advisor: Pradeep Radhakrishnan, PhD
Led the mechanical redesign, perception and planning, actuation, and human-robot interaction for an open-source 3D-printed humanoid built for assistive healthcare, using a 9-DOF IMU and kinematic trajectory planning for locomotion and physics-based digital-twin simulation to resolve sim-to-real divergence in mobile manipulation. In simulation, the robot sustained unassisted standing for over 3.5 hours, squatting for over an hour, and cart-pushing for over an hour; deployed on hardware, it achieved 30 minutes of unassisted standing and pushed a cart over 2 meters, surfacing the real sim-to-real gap.
TurtleBot SLAM
Fall 2023
Advisor: Greg Lewin, PhD
Autonomous navigation system for a TurtleBot3 mobile robot implementing differential-drive kinematics, probabilistic state estimation (Kalman and particle filters), and optimal path planning.