Precision agriculture for corn using reinforcement learning

Tan, Carlton (2024) Precision agriculture for corn using reinforcement learning. Final Year Project, UTAR.

Preview

PDF Download (3845Kb) | Preview

Abstract

This project introduces RACKY, an innovative and comprehensive API solution that seamlessly integrates the SWAT (Soil and Water Assessment Tool) model into the highly adaptable OpenAI Gym environment, thus creating the powerful simulation framework known as SWATGym. RACKY serves as a versatile interface, facilitating effortless retrieval of detailed corn plant state information based on precise fertilizer or irrigation inputs through its intuitive API endpoints. Beyond data access, RACKY incorporates a sophisticated reinforcement learning agent based on the Proximal Policy Optimization (PPO) algorithm within the SWATGym. This integration empowers users with the capability to input location-specific data alongside plant growth stage parameters, thereby obtaining highly optimized recommendations for fertilizer and irrigation amounts directly from the embedded AI model. RACKY helps people make better farming decisions by showing them how different amounts of fertilizer and water affect plant growth through real-time simulations and detailed analysis. This project aims to make advanced farming information and AI tools accessible to everyone, not just experts. By using RACKY, researchers, farmers, and anyone interested in farming can find ways to grow crops more sustainably and using fewer resources.

Actions (login required)