In a recent article published in the journal Robotics, researchers presented an innovative approach to automating the inspection of blast holes in open-pit mining environments, a task traditionally carried out manually.
Image Credit: Pajor Pawel/Shutterstock.com
Background
In open-pit mining, the process begins after drilling and prior to blasting, where the blast holes – vertical or near-vertical openings drilled into the earth – are filled with explosives. These holes typically span approximately 10 meters in depth and have diameters ranging from 20 to 40 centimeters.
The quality of the drilling and the precision with which the holes are inspected plays a vital role in ensuring effective blast outcomes and controlling fragmentation. Traditionally, human operators perform visual and manual inspections, often assisted by manual tools or remote-controlled devices. However, these procedures are limited in accuracy and can be hazardous under certain conditions.
Previous research has focused on enhancing mine automation via robotics, perception, and control systems, but the focus has often been on surface-level navigation, mapping, and obstacle avoidance, rather than specific underground or borehole inspection.
In particular, efforts to develop autonomous systems for underground mining face challenges related to the deployment of sensing technologies in confined, dusty, and unpredictable environments. The use of laser scanners (LiDAR), computer vision, and sensor fusion techniques has increased in recent years, aiming to enable robots to perceive and understand complex mine environments with high accuracy.
The Curent Study
The core methodology revolves around integrating perception and navigation to achieve precise inspection and sensor insertion into blast holes. The robot, DIPPeR, is an autonomous ground vehicle equipped with sensors including LiDARs, GPS, IMUs, wheel encoders, and a dedicated dipping actuator for sensor insertion.
The navigation strategy is proximity-based and adaptive, designed to compensate for the deficiencies of GPS signals often encountered in open-pit mines. Particularly, as the robot approaches a target blast hole, it switches from relying solely on global positioning to local relative localization. When the target is detected visually, the robot dynamically adjusts its guidance parameters to maintain optimal proximity and alignment for inspection.
Perception is primarily achieved through processing LiDAR data. The process begins by collecting 3D point cloud data, which is then segmented to isolate ground and above-ground features. Given that blast holes are identified by the shape of a cone of excavated or waste material above the bore, the system focuses on extracting these cone-shaped volumes from the 3D point cloud. This involves projecting the 3D point cloud into a virtual depth image, forming an accurate 2D segmentation. Within this 2D image, the algorithm employs a detection pipeline based on a coarse-to-fine approach. This includes initial candidate region extraction, followed by refined circle fitting using least squares algorithms to identify the precise centroid and radius of the detected hole.
Results and Discussion
The experimental results demonstrate significant progress in automating blast hole inspection in surface mining environments. The initial simulations verified the feasibility of the perception algorithms, with successful detection of virtual blast holes at varying distances and angles. Simulated tests allowed for parameter tuning and validation of the detection pipeline, confirming its robustness to changes in camera projection parameters and target distances.
Moving to controlled laboratory settings, the robot successfully identified holes with varying cone geometries. These tests confirmed the capability of the circle-fitting algorithms to localize boreholes with high accuracy, typically within a few centimeters, which is sufficient for precise sensor placement. The non-maximum suppression process contributed to reducing false detections and overlapping candidates, leading to reliable target identification.
The real mine-site trials were the most telling in terms of practical applicability. The robot navigated safely around mine slopes and debris, compensated for noisy GPS signals, and relied on the perception pipeline to find, track, and localize the blast holes. The adaptive navigation approach allowed the robot to maintain close proximity to targets without collision, despite irregularities in terrain and unexpected obstacles. The system achieved high detection rates, with minimal false positives, and successfully positioned the sensor probe within the boreholes for data collection or sensor insertion tasks.
Conclusion
The study emphasizes the significance of robotic automation in contemporary mining operations, particularly in the critical task of blast hole inspection. The results from simulations and real-world field tests confirm that the approach is capable of reliably detecting and localizing blast holes, even in the face of noisy GPS signals and environmental variability.
The system’s capability to maintain continuous tracking and precise sensor insertion positioning has the potential to transform blast hole inspection from a labor-intensive manual process to an automated, safe, and efficient operation. This transition promises improved safety for personnel, quicker inspection turnaround times, and potentially significant cost savings for mining companies.
Journal Reference
Liu L., Mihankhah E., et al. (2025). Blast Hole Seeking and Dipping: Navigation and Perception Framework in a Mine Site Inspection Robot. Robotics 14(12):173. DOI: 10.3390/robotics14120173, https://www.mdpi.com/2218-6581/14/12/173