Strategic bidding behaviour in agricultural land rental markets: reinforcement learning in an agent-based model

Abstract Agricultural land markets are crucial for efficient land allocation, yet they face complexities arising from land characteristics and the heterogeneous nature of market participants. This study explores how to address heterogeneity in the modelling process for land markets models by integrating Deep Reinforcement Learning (DRL) into the agent-based model AgriPoliS, to model strategic bidding behaviour. The simulations demonstrates that a DRL agent adapts its bidding strategies based on long-term growth objectives, experience, competitive interactions and adaptive decision-making leading to increased land rental and farm growth compared to a standard agent using a fixed bidding strategy. The results reveal how strategic behaviour not only improve individual farm performance but also affect neighbouring farms, emphasizing the dynamic interactions within land markets. By capturing the agent’s strategic behaviour, this work contributes towards more realistic modelling of agricultural land market dynamics and offers insights into the implications of potential land market regulations. Future research will explore multi-agent frameworks to further refine these interactions and address the limitations of static bidding strategies.