Hyperparameter Configurations for Greenhouse RL Algorithms

Hyperparameter settings for reinforcement learning algorithms tested in a greenhouse climate control simulation. The dataset supports the evaluation of Proximal Policy Optimization, Deep Deterministic Policy Gradient, Soft Actor-Critic, and Twin Delayed Deep Deterministic Policy Gradient algorithms. It originates from a study by Iman Hindi, published on figshare in 2026.

Use Cases

• Compare the performance of PPO, DDPG, SAC, and TD3 algorithms using the documented hyperparameter configurations.
• Reproduce the greenhouse control simulation environment that uses MLP and LSTM models for climate and crop prediction.
• Analyze the custom reward function design based on stem elongation, stem thickness, cumulative trusses, and resource consumption costs.
• Benchmark RL algorithm convergence speed and cumulative reward outcomes against the reported TD3 results.

Strengths

• Dataset is openly licensed under CC BY 4.0.
• Explicitly compares four RL algorithms: PPO, DDPG, SAC, and TD3.
• Focuses on a defined application: greenhouse climate control for crop yield and resource optimization.

Limitations

• Very small file size (5.5 KB) indicates limited data volume, likely containing only configuration parameters.
• No sample data or column definitions are provided, limiting immediate analytical utility.
• Dataset is derived from a simulation study, not real-world operational data.

Provenance

Source

figshare

Collection Method

Generated from a reinforcement learning simulation study for greenhouse climate control.

Time Range

null

Freshness

Last updated in March 2026.

Geography

null