Material grade-based task planning for robotic limestone mining

This paper focuses on task planning for the robotization of wheel loaders in limestone mining to address growing material demands and workforce shortages by boosting the production rate. Wheel loaders repeatedly scoop and unload raw materials into trucks at extraction sites. Sustaining consistent and high production rates requires effective grade management and high vehicle operational efficiency. However, existing studies have not integrated material grade management into loading operations, a critical aspect of limestone mining. This paper addresses the limitation by proposing a novel task planning approach based on agent-centered search, called the Best First Search with Grade Estimation (BFSwGE) algorithm. The algorithm aims to maximize the extraction of high-grade material while minimizing vehicle travel distances during loading operations. It optimizes the sequence of the wheel loader's scooping poses on the material pile and unloading poses at which the vehicle dumps the material to a truck. By leveraging hyperspectral imaging to assess the material grade distributions of the pile surface, the algorithm estimates the material grade distribution of the pile volume using Laplace Interpolation, enabling a more effective search. Simulation results across various pile grade distributions demonstrate that the proposed algorithm outperforms a greedy search method in achieving these objectives.