Evaluating Underground Mining Automation Platforms

Underground automation used to mean a loader that could tram a heading while the operator sat in a cabin on surface. That is still the job. The software question has moved. Vendors now sell a platform: navigation, traffic, access control, and a picture of the whole mine in one stack.

In May 2026 Sandvik launched AutoMine Aura. In early September it was on the floor at Future of Mining in Tampere. Canadian Mining Journal covered the showcase. That is the in-market trigger. It is not a reason to pick a winner. It is a reason to write down what an underground automation lead should verify before an upgrade, a trial, or a first buy.

MSR lists an independent profile for AutoMine. There is also an autonomous mining category. Neither is a ranking.

For twenty years AutoMine, according to Sandvik, has been a production automation system in both underground and surface mines. Aura, according to the same vendor, is not the next version. It is a new platform. Sandvik says it launches first on underground loaders and will spread to other Sandvik machines later.

The claims to label, not to swallow:

• Sandvik describes Aura as a first-of-its-kind underground automation platform.
• The company says a new 3D perception navigation system has been proven to move over 15 percent more material at one harsh underground mine.
• David Hallett, Sandvik vice president of automation, said AutoMine has been used at more than 140 mines, and that Aura was built as an entirely new platform rather than a revision.
• Sandvik says the platform integrates with existing networks and access control, so mines do not have to rip out infrastructure.
• The company also cites almost nine million hours without a lost-time injury on AutoMine. That is a safety statistic from the vendor. It is not your site's risk assessment.

What you can take from that without believing the adjectives: OEM automation is moving from "this machine can tram" to "this software tries to see the heading, the traffic, and the next loader." If you already run AutoMine, the live question is upgrade path. If you do not, the live question is whether you are buying a machine feature or a mine operating system.

A loader that can tram a drawpoint is a machine problem. A platform that coordinates several machines across levels is a mine problem. They get sold as the same slide.

Mine-wide means traffic rules, access control, people in the heading, and a second OEM's gear that still has to move. If Aura, or any other OEM platform, only sees Sandvik loaders, you do not have mine-wide automation. You have a gated paddock.

Ask where the traffic brain lives. If it lives in the OEM cloud or in a box that only talks to one fleet, a second vendor's truck becomes a guest that needs a escort. That may be acceptable on a single-OEM site. It is a different decision on a mixed fleet.

Ask what "3D perception" needs in the heading. Extra sensors, extra compute, extra people who can keep those sensors clean in mud. A demo in Finland is not a development heading after blast.

Ask whether the operator interface is supervision of many machines or remote control of one. Those are different roster and fatigue problems. A wider screen does not tell you how many machines one person can actually watch when two headings bog at once.

1. Fleet mix. Does the platform traffic your current loaders and trucks, or only the OEM's new units? Get the list in writing.

2. Start small. Sandvik says Aura starts on underground loaders. If that is the offer, the trial should be one level, one machine class, measured against your current cycle time, not a site-wide promise.

3. Network. "Works on existing networks" is a vendor claim. Have your OT person say which VLAN, which latency, and what happens when fibre takes a hit.

4. Access control. People still walk. Confirm how pedestrians, light vehicles, and the automated zone interact, and who can override.

5. Data. Where do tram logs, perception logs, and incident records live? Can you export them without a vendor ticket? If the answer is no, you do not own the evidence when someone asks what the machine saw.

6. Hardware service. Who cleans, calibrates, and replaces the 3D sensors on night shift? If the answer is a fly-in specialist, price that into availability.

7. Upgrade versus rip. Existing AutoMine sites should ask what of the old install survives. A new platform that needs new cabins, new computers, and new training is a project, not a patch.

8. Exit. If you stop paying, do the machines still tram in manual, and do you still have the logs. Lock-in is not evil. Unspoken lock-in is.

None of those questions name a winner. They stop you buying a keynote.

Timeline, for the file: launch announcement 26 May 2026, industry showcase 1-3 September 2026 in Tampere. Coverage in International Mining and Canadian Mining Journal. Scanner signal on 3 September 2026.

Treat every productivity and safety number as Sandvik's number until a trial on your ground says otherwise. "Zero blind spots" is marketing language. Underground always has blind spots. Your job is to find where they moved.

If you evaluate Aura, run the checklist on one heading you actually own. Bring last month's tram times, last month's bogging events, and the person who currently sits in the remote cabin. If the demo cannot use that heading's map, you are watching a video.

A fair trial is boring on purpose. Pick one loader, one level, two weeks. Measure cycle time, hang time, and how often a person had to take over. Count the hours the sensors were dirty or the network was sick. If the vendor wants a site-wide scoreboard in week one, they are selling a keynote.

Existing AutoMine sites have a second job: map what still works. Cabins, access control, traffic lights, and the people who already know the old screens. A platform change that throws that away is a shutdown, even if the slide says "no infrastructure changes." Have the OT lead and the remote operator in the same room when that sentence is said.

Do not turn this into a beauty contest between OEM brands. The useful split is machine feature versus mine operating system, single-OEM paddock versus mixed traffic, and whether you can still produce when the vendor is not on the plane.

Automation software sits next to fleet management and dispatch, not on top of it. Dispatch still has to know what the automated zone is doing. If the two clocks disagree, the shift boss will believe the one that matches the orepass.

It also sits next to people. Remote supervision changes who is underground and who is in a chair. That is a roster and a competence problem, not a feature tick.

The directory's autonomous mining list is the place to see what is listed. This article is the questions to take into a trial. The pit will not care that the platform had a launch in Finland. It will care whether the loader still trams when the heading is full of fog and the network is sulking.