Por: Alem Huayta Uribe, Charly Cueva Ricse y Diego Cajachagua Guerreros, Universidad Continental.

Abstrac

This technical study details the comprehensive development of an autonomous robotic system designed specifically for geological exploration and sampling in underground mines. The system is engineered to operate in GPS-denied environments and under adverse environmental conditions. 

The proposed solution integrates SLAM-based navigation algorithms, multispectral perception (RGB + IR), articulated robotic manipulation, and an automatic return system, all built on a high-efficiency embedded architecture. Validation was conducted in four sequential phases: autonomous path planning, functional simulation in underground environments, subsystem-level energy modeling, and analysis of sampling time accuracy. Simulations demonstrated stable energy consumption (62 Wh/h), with power consumption concentrated in movement (50%), sensing, and manipulation, while maintaining a mean SLAM localization error of 3.82 cm. 

Sampling times ranging from 8.4 to 18.5 seconds per event were also recorded, depending on the structural complexity and optical conditions of the environment. Multispectral perception enabled the identification of geologically relevant areas through thermal and spectral contrasts, improving sampling efficiency. 

Overall, the prototype represents an advanced tool for autonomous underground mining operations, combining operational precision, energy autonomy, and adaptive intelligent navigation. It also has the potential to be scaled for geotechnical reconnaissance and monitoring missions in confined or hard-to-access spaces.