Precision Agriculture Optimization based on Multi-Armed Bandits Algorithm: Wheat Yield Optimization under Different Temperature and Precipitation Conditions
Qikang Huang · ITM Web of Conferences · 2025
Climate change and the growing unpredictability of environmental elements such as temperature and precipitation present considerable challenges to contemporary agriculture. Data-driven algorithms present promising solutions by offering more precise tools for optimizing crop yields and resource efficiency to tackle these challenges. Among these approaches, the multi-armed bandit (MAB) algorithm effectively balances exploration and exploitation, showcasing considerable potential for optimizing agricultural decision-making. This study investigates four widely utilized Multi-Armed Bandits (MAB) algorithms: Explore Then Commit (ETC), Upper Confidence Bound (UCB), Asymptotically Optimal UCB, and Thompson Sampling (TS). The objective is to optimize wheat yield under varying temperature and precipitation conditions while also assessing the effectiveness of these different algorithms in achieving this goal. The experiment demonstrates that the UCB algorithm is optimal for analyzing data on total precipitation during the growth of wheat. . Furthermore, the TS algorithm significantly outperforms others in analyzing flat temperature data throughout the wheat growth period. Therefore, the Asymptotically Optimal UCB algorithm can identify the most suitable rainfall conditions for wheat growth in a changing environment. In contrast, the TS algorithm can determine the optimal temperature requirements for wheat growth under similar environmental fluctuations. These insights assist agricultural practitioners in timely adjusting their strategies to enhance crop yield. Additionally, it provides a model for those who want to use the MAB algorithm to improve agricultural yields.