Abstract
In underground mining stoping methods, the stope is the void space created when ore is excavated from the surrounding rock. The set of stopes that make up the layout is subject to optimization and has been the focus of various works. This study addresses stope layout optimization through reinforcement learning (RL). The proposed method consists of two stages. The first stage is preprocessing, where stopes are created from geometric parameters and available information in the block model. The second stage is optimization, where a reinforcement learning algorithm is applied, allowing an agent to learn to select stopes to maximize the economic value of the final layout. The application of this method to a case study demonstrates that reinforcement learning finds solutions efficiently. Furthermore, the use of reinforcement learning in mine planning offers significant potential to optimize processes, improve decision-making efficiency, and dynamically adapt to changing environmental conditions.