02 / GOOGLE DEVELOPER STUDENT CLUB / MARCH 2026 – MAY 2026Lunar
Lunar
Lander.
A reinforcement learning project built with a seven-person team. I focused on training and evaluating the agent as it practiced autonomous landings.
Watch the demo
ROLESoftware Developer — ML Training & Evaluation
TEAM7 members
PERIODMarch 2026 – May 2026
PROJECT OVERVIEW
Learning to land.
The agent learns to steer a lunar module toward a safe landing between two flags. Our team used repeated training runs and reward tracking to assess its progress.
What I contributed
- Trained and evaluated the agent across 2,000 episodes.
- Logged rewards and saved model checkpoints every 250 episodes to compare progress.
- Used Matplotlib plots to inspect trends and identify crash penalties.
- Adjusted target-network synchronization to stabilize learning after early negative scores.
PROJECT DEMO
See the lander in action.
The team presentation opens on the demo slide. Use the video controls inside the slide to play the landing run.
SOURCE CODE
Open Lunar Lander on GitHub Explore the project.
See the agent, environment setup, and visualization code.
