Abstract
We introduce WOFOSTGym, a novel crop simulation environment designed to trainreinforcement learning (RL) agents to optimize agromanagement decisions forannual and perennial crops in single and multi-farm settings. Effective cropmanagement requires optimizing yield and economic returns while minimizingenvironmental impact, a complex sequential decision-making problem well suitedfor RL. However, the lack of simulators for perennial crops in multi-farmcontexts has hindered RL applications in this domain. Existing crop simulatorsalso do not support multiple annual crops. WOFOSTGym addresses these gaps bysupporting 23 annual crops and two perennial crops, enabling RL agents to learndiverse agromanagement strategies in multi-year, multi-crop, and multi-farmsettings. Our simulator offers a suite of challenging tasks for learning underpartial observability, non-Markovian dynamics, and delayed feedback.WOFOSTGym's standard RL interface allows researchers without agriculturalexpertise to explore a wide range of agromanagement problems. Our experimentsdemonstrate the learned behaviors across various crop varieties and soil types,highlighting WOFOSTGym's potential for advancing RL-driven decision support inagriculture.