Tang et al. (2026) Research on Adaptive Irrigation Decision‐Making Method for the Entire Growth Cycle of Water Spinach Based on Reinforcement Learning

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Food Frontiers
• Year: 2026
• Date: 2026-03-30
• Authors: Ran Tang, Loy Chee Luen, Wei Sun, Narendra Kumar Aridas, Mohamad Sofian Abu Talip
• DOI: 10.1002/fft2.70261

Research Groups

Not specified in the abstract.

Short Summary

This study proposes an environmentally enhanced proximal policy optimization (EN-PPO) method for precision irrigation control in water spinach production, which addresses challenges from rainfall uncertainty and crop growth stage differences by incorporating a dynamic shearing strategy and a negative incentive mechanism, demonstrating superior performance in water saving, rainfall utilization, and stable policy convergence without affecting crop yield.

Objective

• To develop a precision irrigation control method for water spinach production that can effectively manage rainfall uncertainty and crop growth stage differences, moving beyond traditional experience-dependent irrigation decisions.

Study Configuration

• Spatial Scale: Field/Plot scale for water spinach cultivation.
• Temporal Scale: Growing season of water spinach, encompassing multiple irrigation cycles and training phases.

Methodology and Data

• Models used: Environmentally enhanced proximal policy optimization (EN-PPO), built upon the traditional Proximal Policy Optimization (PPO) reinforcement learning framework.
• Data sources: Environmental information (e.g., rainfall, soil moisture levels) and crop growth stage data, likely from sensors or simulations, used for training and evaluation of the irrigation policy.

Main Results

• The EN-PPO algorithm exhibits superior comprehensive performance in policy convergence stability, water-saving effect, and rainfall utilization efficiency.
• It achieves more reasonable irrigation timing and water volume regulation without negatively affecting normal crop growth and yield.
• The introduced dynamic shearing strategy effectively mitigates training oscillations caused by rainfall randomness and sample scarcity.
• The designed negative incentive mechanism successfully guides the agent to avoid high-risk decisions during exploration and maintain stable, water-saving irrigation strategies during utilization, by penalizing violations of soil moisture safety ranges, redundant irrigation, and unmet rotation/switching constraints.

Contributions

• Proposes EN-PPO, an innovative reinforcement learning framework specifically designed for precision agricultural irrigation, enhancing traditional PPO with environmental information.
• Introduces a dynamic shearing strategy for adaptive policy updates, significantly improving training stability under strong agricultural uncertainties like rainfall randomness and sample scarcity.
• Develops a negative incentive mechanism to ensure production safety and resource efficiency, guiding the irrigation agent towards safe and water-saving decisions.
• Provides a feasible and robust approach for the engineering application of reinforcement learning in agricultural precision irrigation, addressing practical challenges.

Funding

Not specified in the abstract.

Citation

@article{Tang2026Research,
  author = {Tang, Ran and Luen, Loy Chee and Sun, Wei and Aridas, Narendra Kumar and Talip, Mohamad Sofian Abu},
  title = {Research on Adaptive Irrigation Decision‐Making Method for the Entire Growth Cycle of Water Spinach Based on Reinforcement Learning},
  journal = {Food Frontiers},
  year = {2026},
  doi = {10.1002/fft2.70261},
  url = {https://doi.org/10.1002/fft2.70261}
}