Underground mining accidents are rarely random. They tend to occur in predictable conditions: at intersections, around blind corners and in confined drifts where vehicles and people move through shared space without full visibility. In these environments, operators are required to make decisions with incomplete information and within limited reaction windows.
The interaction between heavy mobile equipment and personnel is often compressed into seconds, leaving little margin for error once a safety hazard becomes visible. This constraint – limited visibility in a dynamic operating environment – remains central to how accidents occur underground.
It is not simply a matter of poor lighting or operator awareness but a structural condition shaped by tunnel geometry, dust and the physical limits of human perception. In practice, the most dangerous interactions are often those that cannot be seen until they are already unfolding.
Improving safety metrics, but not uniformly
At an industry level, safety performance is improving. The ‘Safety Trends in Mining 2025‘ report by Mining Technology’s parent company, GlobalData, shows that the average number of fatalities across 54 mining companies declined from 4.3 in 2023 to 3.1 in 2024. Injury rates have also fallen, with the total recordable injury frequency rate decreasing from 2.87 to 2.64 and the lost-time injury frequency rate was reported at1.55. These figures indicate measurable progress across the sector.
However, the drivers behind these improvements are mixed. GlobalData highlights that part of the reduction in accidents reflects structural changes in company portfolios rather than uniform improvements in operating conditions. ArcelorMittal’s decline in fatalities follows the sale of its Kazakhstan operations after a fatal explosion that killed 46 workers in October 2023 at the Kostenko coal mine, effectively removing a high-risk asset from its reporting base. Elsewhere, improvements at Impala Platinum are linked to site-level safety reforms.
At the same time, performance remains uneven across operators. Some companies recorded zero fatalities over multiple years, while others continue to report significantly higher incident counts. In 2024 alone, 16 companies reported no fatalities while others recorded multiple deaths, including 30 at Coal India. Injury rates show a similar spread.
The data points to improvement, but an uneven one. Safety outcomes are moving in a positive direction overall, but they are not doing so consistently across the industry, and they do not necessarily reflect a fundamental shift in the conditions that generate risk underground.
Why accidents persist in underground mining
The persistence of accidents underground is closely tied to the limits of conventional safety systems.
Most collision avoidance technologies currently deployed in mining rely on line-of-sight detection. Cameras, radar and LiDAR identify hazards based on what can be seen. In open-pit environments, where visibility extends over long distances, this approach is effective. Underground, those assumptions break down.
Tunnel geometry creates blind zones at intersections and around bends. Dust and moisture degrade sensor performance and reduce clarity. Even under optimal conditions, the detection range is constrained by what is physically visible in front of the vehicle. At typical underground operating speeds, this significantly reduces the time available for operators to recognise hazards and respond.
Research into underground haulage systems shows that detection distance must exceed stopping distance to be effective under operating conditions. Where that condition is not met, systems may still detect hazards accurately, but too late to prevent an incident. The issue is not simply detection, but detection with sufficient lead time to enable braking and avoidance.
In underground settings, this variability is compounded by limited visibility. Operators are not only reacting to dynamic conditions but doing so with incomplete information. The combination of changing conditions and constrained awareness reduces the time available to respond, increasing the likelihood of incidents precisely where coordination between vehicles and personnel is most critical.
Further, surface mining relies on global navigation satellite systems (GNSS), the satellite-based infrastructure behind GPS [global positioning system], to track vehicles, equipment and personnel in real time across large, open environments. GNSS provides a continuous spatial reference, allowing operators and systems to know where assets are and how they are moving.
As GNSS signals do not penetrate rock in underground environments, the result is not simply reduced accuracy. It is the absence of a unified spatial reference system across the operation. Instead of continuous, real-time positioning, mines rely on a patchwork of alternatives, including radio frequency identification (RFID) tags, Wi-Fi triangulation, inertial navigation systems and fixed beacons. Each of these systems provides partial coverage, often limited to specific zones or infrastructure points.
“You don’t have a single source of truth underground,” says Gideon Slabbert, transformation lead at Maptek. “You have partial visibility depending on infrastructure, and that changes across the mine.”
This fragmented awareness creates gaps in situational understanding. Vehicles and personnel may be tracked in certain areas, but not others, and the accuracy of positioning can vary significantly. In a dynamic environment, these inconsistencies reduce the reliability of system-level awareness.
This matters because equipment capability has outpaced system awareness. Underground trucks can operate at speeds where stopping distances exceed what operators or systems can reliably perceive in time. Without a continuous positioning framework, it becomes more difficult to anticipate interactions between moving assets before they enter each other’s field of view.
Technologies improving visibility and connectivity underground
In response, some operators are moving away from purely visual detection systems.
At Rio Tinto’s Oyu Tolgoi mine in Mongolia, a proximity awareness system based on vehicle-to-everything (V2X) communication enables vehicles and personnel to exchange signals directly. Rather than relying on sensors to detect hazards visually, the system allows equipment and workers to broadcast their presence in real time.
“It’s really that distance in non-line-of-sight engagement that we get,” adds Slabbert. “Vehicles can travel at speed and still be aware of each other far enough out to react.”
Each vehicle is equipped with a communication unit, while personnel carry wireless tags integrated into cap lamps. These devices continuously transmit signals, allowing nearby vehicles to detect their presence even when visibility is obstructed by tunnel geometry or environmental conditions.
Signals propagate through the underground environment, extending around corners and through areas where conventional sensors would be ineffective. The system operates without reliance on GNSS and does not depend on continuous fixed infrastructure, allowing it to function in dynamic and evolving mine layouts.
Roobuck and Spectrum FiftyNine played a central role in deploying and integrating the system at Oyu Tolgoi, aligning communications, hardware and operational workflows into a functioning safety layer.
This represents a shift from detection to communication. Instead of attempting to improve visibility, the system extends operational awareness to a new level. Meanwhile, the inclusion of personnel within this network depends on wearable devices.
Workers carry transmitters that allow them to be detected by nearby vehicles, effectively integrating them into the safety system. These devices act as both identifiers and communication nodes, enabling real-time interaction between people and machines. GlobalData’s ‘Wearable Tech (2026)‘ report identifies a broader trend in which wearable technologies are evolving into connected systems capable of collecting and transmitting data in real time. These systems combine sensors, connectivity and processing capabilities, allowing them to function within wider digital environments.
In mining, the application is more focused. Wearables enable proximity detection and real-time alerts, linking workers directly into operational safety networks and extending awareness to include human movement as well as equipment.
“They’re not just tracking devices anymore,” says Slabbert. “They’re part of the system that manages risk.”
A long road ahead to ensure underground safety
The deployment of systems such as those at Oyu Tolgoi demonstrates what is technically possible, but adoption across the industry remains uneven.
Arman Hazrathosseini, researcher at The Robert M Buchan Department of Mining at Queen’s University in Canada, points to several barriers: high initial costs, the difficulty of building communication infrastructure, the lack of specialised staff to run digital systems and resistance from a unionised workforce. Cost constraints are particularly acute for smaller operators, where capital investment and technical capacity may be limited.
GlobalData similarly highlights variation in safety outcomes across operators, suggesting that the implementation and effectiveness of safety systems underground differ widely. Differences in operating conditions, asset portfolios and investment capacity all contribute to this uneven adoption. The technologies now being deployed underground are shaped by the environment in which they operate.
Where visibility is limited, systems extend awareness beyond line of sight. Where positioning systems are unavailable, they rely on local communication networks. Where operations are dynamic, they are designed to function without fixed infrastructure.
These approaches address specific limitations but do not remove them. Even as systems become more connected and responsive, underground mining remains an environment where uncertainty cannot be fully engineered out.
The effectiveness of these systems thus still depends on human response.
V2X systems provide alerts, but they do not automatically intervene or stop vehicles. Operators must interpret signals and act within the available time window. In confined underground environments, that window can be limited.
Hazrathosseini emphasises that mining remains a dynamic environment requiring human oversight: “We need humans to handle unexpected surprises like sudden rock fires or extreme weather… and to act as the final safety kill switch if the technology makes a mistake.”
Experts say this may place a practical limit on what technology alone can achieve. Systems can extend awareness and improve the timing of responses, but they do not remove the need for human judgment, particularly in complex or unforeseen situations.
However, as technology continues to evolve and improve, we may move closer to a world where technology, if not fully guarantees, then nearly promises safety underground.
