Underground communications used to be the part of a mine where the rest of the technology stack went to die. Voice radios that worked sometimes. Tag systems that lost coverage in stope intersections. Data links that pretended to be reliable until the moment you needed them. Most of that has improved. None of it has improved evenly.

In May 2026, the picture is mixed. Some operations are running underground LTE networks with latency comparable to surface fibre. Others are still putting up with mesh systems that drop frames during shift change. The capability gap has implications for everything from autonomous equipment readiness to safety system design.

What good actually looks like

The leading operations I’ve seen running modern underground comms have a few things in common. The backbone is fibre to as many points underground as practical. The wireless layer on top is purpose-built for the environment — typically dedicated underground LTE or private 5G in newer deployments, with Wi-Fi 6 as the fallback for specific zones rather than the primary medium.

Latency on these networks is in the 20-50 millisecond range from underground equipment to the operations centre, including the round trip. That’s good enough for telemetry, voice, video, and most control applications. It is not good enough for true real-time control of equipment with tight closed loops, but those use cases are not really live in production yet.

The problem zones in the leading operations are predictable. Areas just commissioned have less infrastructure in place. Areas that have been mined out have legacy equipment that needs to be repurposed or removed. Areas with heavy pillar load have signal propagation issues that are hard to design around.

Where most operations actually sit

Most Australian underground operations sit somewhere in the middle. There’s an LTE or mesh network in the active development zones. There’s some fibre. There’s Wi-Fi in fixed installations. The latency profile is highly variable — measured in milliseconds in some locations and in seconds in others, depending on what the user is doing and where they are.

The variability is the practical issue. A network with consistent 100ms latency is more useful than a network that’s 30ms most of the time and 800ms when a particular cell is overloaded. Applications that need to function in the worst case have to be designed for the worst case.

This is why some of the more ambitious applications — fully autonomous loaders running tight cycle times, real-time geotechnical monitoring with closed-loop response — remain in pilot rather than production. The comms can support them most of the time. They cannot yet support them all the time.

The applications that are gated by latency

Three application categories are most tightly gated by underground comms latency in May 2026.

Tele-remote and autonomous equipment is the obvious one. Tele-remote loaders work fine on networks with sub-100ms latency and reliable bandwidth. Autonomous loaders with full closed-loop control of speed and steering need tighter latency and stronger reliability guarantees. The autonomous fleet sizes are creeping up, but the deployment pattern reflects what the comms can support.

Real-time monitoring with response loops is the second. Slope monitoring, ventilation-on-demand, ground support monitoring — all of these can be built as latency-tolerant systems where the underground sensor logs locally and uploads when a connection is available, or as low-latency systems where the operations centre responds in real time. The latency-tolerant version is more common because it works with the comms most operations actually have.

Voice and video collaboration between underground crews and surface specialists is the third. The use case is mature and the demand is real. The experience of trying to do real-time fault diagnosis over a video link with intermittent latency is bad enough that most crews fall back to voice or to physically bringing the surface specialist underground. Operations with reliable underground comms get more out of their specialist labour.

The standards conversation

There is no single Australian standard for underground comms performance. There are vendor specs, operational targets that individual operations set internally, and the de facto patterns that emerge from what actually works.

The Australian Centre for Robotics and the relevant industry bodies have been pushing for clearer specifications, particularly around latency and reliability targets that should underpin autonomous operations. Progress has been incremental. The vendors have a commercial interest in interoperability that hasn’t fully translated into common specifications.

For operations planning new networks or major upgrades, the practical advice from peers is to design for the autonomous use case you expect three years out, not the use case you have now. Pulling new infrastructure underground is expensive enough that an underspecified network will be painful to upgrade. An overspecified network is uncomfortable budget-wise but useful operationally.

Where this goes next

By the end of 2026 I expect more operations will have private 5G as the dedicated underground network. The advantages over LTE are real for high-density telemetry and for the autonomous use cases. The cost has come down enough that it’s a reasonable choice for new builds.

I also expect the gap between the leading operations and the average operation will widen. The leading operations are compounding both their physical comms infrastructure and the institutional knowledge of how to design applications for the environment. The average operation is still upgrading from the previous generation of network.

The lesson from looking across the spectrum is that comms infrastructure is the thing that gates everything else underground. Get that right and the rest of the technology stack has somewhere to live.