267 The following table illustrates the acronyms used throughout this summary: Table 1 Acronyms Code Meaning V (NPV) Net Present Value IA Artificial Intelligence RL Reinforcement Learning R Mining Resources C, c Operating Cash Flow g Policy of each period of extraction F Opportunity Cost \ Rate of time for unit of resource s Current State of Systems r Reward a Action P Probability ] Learning Parameter MILP Mixer Integer Lineal Programming ^ Set of polices MDP Markov Decision Process LLM Large Language Model NP Non-Polynomial LOM Life of Mine GRASP Metaheuristic Algorithm TABU Metaheuristic Algorithm 1.1. Problem Statement Current mining operations face a global challenge: delivering essential minerals faster and more efficiently while maintaining economic viability. The 'decision gap' in strategic planning—caused by the latency of traditional optimization models—results in sub-optimal resource utilization. This research addresses this gap by integrating Lane’s fundamental economic theory with the computational power of Reinforcement Learning, enabling real-time policy adjustments that traditional MILP (Mixed-Integer Linear Programming) models cannot achieve under high uncertainty. 2. METHODOLOGY The methodology begins with the development of a mathematical model based on Lane’s Theory (Lane,1988), which represents the geological and operational constraints of the mine, such as processing and mining capacities and serve like modeled environment. In parallel, a RL agent is trained to interact with this modeled environment. The agent receives rewards based on the economic and operational impact of its decisions, allowing it to gradually improve the production plan through the training process. To better understand, we will briefly explain each finding. First, Kane Lane (Lane,1988) propose the Equation 1. The purpose of the equation is to determine the maximum Net Present Value (NPV) of the mining project and optimal cut‑off grade in the R space.
RkJQdWJsaXNoZXIy MTM0Mzk2