Autonomous Rock Detection for LHD Vehicles in Underground Block Caving

Underground mining, a common extraction method for rocks, presents obstacles in the removal and transportation of materials phase, like limited space, strict safety requirements, and high energy consumption, which can account for a substantial portion of operational costs, leading to congestion and delays in underground roads. To address these challenges, this research focuses on the development of an autonomous Load-Haul-Dump (LHD) vehicle aimed at enhancing operational efficiency, safety, and productivity in underground mining operations. This vehicle is equipped with an AI-powered visual inspection system capable of effectively detecting rocks to facilitate the block caving technique. To support the effectiveness of this system, a vibration analysis was performed, which determines displacements of 1.44 um and 1.61 um against two different types of floors. Utilizing the YOLOv9 architecture and having tested with seven different optimizers to find the ideal machine learning parameters for the rocks, the system achieved an F1-Score of 99.12% for rocks and 74.33% for scattered rocks, both optimized with the Stochastic Gradient Descent (SGD) algorithm.