Algae Growth RL Prediction

Python research sandbox for simulating algae growth and training a reinforcement-learning controller. The custom Gymnasium environment models light, nutrient level, temperature drift, ultrasound exposure, and trace elements; a PPO agent can learn action settings that maximize simulated growth.

Project Structure

• env/algae_env.py: Gymnasium environment used by training and tests.
• agents/train_ppo.py: PPO training script using Stable-Baselines3.
• utils/plot_utils.py: plotting helper for growth and factor histories.
• tests/test_env.py: manual environment smoke test.
• quick_algae_report.csv, algae_result.png: generated experiment outputs kept as reference artifacts.

Setup

python3 -m venv .venv
.venv/bin/python -m pip install -r requirements.txt

On Windows, use .venv\Scripts\python.exe instead of .venv/bin/python.

Run

Train and plot a quick PPO run:

.venv/bin/python agents/train_ppo.py

Run the environment smoke test:

.venv/bin/python tests/test_env.py

Notes

• Training requires torch, gymnasium, and stable-baselines3.
• The growth model is a simplified simulation, not a validated biological model.
• Keep large local virtual environments and regenerated scratch outputs out of git.