A Reinforcement Learning approach to study climbing plant behaviour.

Area of Science:

• Plant biology
• Computational biology
• Biomechanics

Background:

• Plant structure results from adaptation and evolution.
• Understanding plant organ mechanics is crucial for biology.

Purpose of the Study:

• Investigate mass and radius distribution in searcher shoots.
• Test the hypothesis that plants maximize length while avoiding stress thresholds.
• Develop a Reinforcement Learning (RL) environment for studying plant mechanics.

Main Methods:

• Created a Reinforcement Learning (RL) environment named Searcher-Shoot.
• Modeled shoot and leaf mechanics, including mass effects.
• Formulated the problem as a Markov Decision Process to analyze mass distribution efficiency.

Main Results:

• Discovered that plant shoots smoothly decrease their diameters along the stem.
• Identified an efficient mass distribution strategy in plant growth.
• Simulated plant behavior using RL, showing good agreement with experimental data.

Conclusions:

• The RL approach effectively mimics and analyzes plant growth traits.
• Efficient mass distribution is key to plant length maximization and stress avoidance.
• This computational method offers a powerful tool for biological systems analysis.