Projects¶

Self-driving algorithm with TD3 for TurtleBot3¶

Master’s thesis · Universidad de los Andes · 2022 — 2025

Designed and trained a reinforcement-learning policy that drives a TurtleBot3 mobile robot autonomously in the Gazebo simulation environment.

Highlights

TD3 (Twin Delayed Deep Deterministic Policy Gradient) for continuous control of the robot’s linear and angular velocities.
Integrated ROS2 topics for sensor input (LiDAR, odometry) and actuation.
Iterated on reward shaping and observation design to handle obstacle avoidance and goal-reaching in cluttered environments.
End-to-end pipeline: simulation setup, training loop, model evaluation, and policy deployment back into Gazebo.

I plan to write up additional projects here — including some of the open work and side experiments around web tooling and AI-assisted development.