TRUCKS nearly as tall as three-storey buildings are a common sight rumbling through the dusty, red-hued landscape of the Pilbara region in Australia.
These haulers, each of which can carry up to 500 tonnes of iron ore, scarcely seem like a tech revolution in the making. But look carefully at some of them as they twist their way up the unpaved mining roads to reach the ore-crushing facilities: the cabs are empty of human occupants.
The trucks are part of mining giant Rio Tinto’s Mine of the Future initiative, which is based at the remote West Angelas iron mine. The firm is betting $500 million that robots are the future of mining.
Advertisement
Rio is far from alone – mining companies from Chile to South Africa to Scandinavia are getting in on the act, too. As the global population continues to climb, the demand for resources including coal, iron and gold is growing at a furious pace. The only way to meet that demand, the reasoning goes, is to turn over the dirty, back-breaking and often perilous job of mining raw materials to artificially intelligent systems that will increase production while making the industry safer for people.
South Africa’s famously deep gold and platinum mines are still mostly worked by people. Each day after blasting is complete, a foreman descends into the fresh tunnels, tapping the roof and listening for a hollow thud that could indicate a hanging wall is in danger of collapsing.
But entry inspections are often rushed, because miners’ pay is tied to hitting benchmarks in a timely fashion. Some in such “ground falls”, according to the country’s Department of Mineral Resources.
“The guys don’t do the entry inspections properly, we know that, and there are fatalities as a result of this,” says of the Council for Scientific and Industrial Research in Auckland Park, Johannesburg, South Africa.
To keep miners out of harm’s way, Vogt and colleagues have built a robot that can navigate the 1-metre-high tunnels on tank-like treads and scan rock faces for weaknesses with a thermal camera. “Rock that is firmly attached to its surroundings will cool more slowly than rock that is broken,” says Vogt.
When weaknesses are spotted, the robot can tap on them with a long arm, and use microphones to listen to the sound it makes. On-board neural network software trained by mine inspectors will recognise if the area is safe. If it seems dangerous, the robot will tag the hanging wall with spray paint, so human crews can bolt the rock or secure it with supports.
Small is big
The team’s robot, which they plan to demonstrate in March next year, is just the beginning, though.
“Our vision is of a fleet of small robots doing various tasks,” including mining and hauling ore, Vogt says.
South Africa’s gold production from many of the easy-to-reach big veins, less than 3.5 kilometres below the surface, is dwindling. There is much more gold deeper down, which companies like AngloGold Ashanti are planning to access with their own autonomous systems. But Vogt says the real prize is in shallow veins 10 to 20 centimetres thick, much too narrow for humans to mine economically. Teams of 10 little robots, say, could drill the ore or blast it with pulses of electricity, then haul back 1 to 2 litres of rock at a time into tunnels accessible by miners. That vision is still a decade away, Vogt admits, barring a breakthrough from one of the many private firms thought to be working on the problem, who remain tight-lipped on the subject.
Where some are thinking small, other mining outfits are already getting on with very big plans.
Swedish firm Sandvik, for example, has deployed in several mines around the world over the last few years.
Beneath the surface, autonomy takes on a new layer of complexity. Out of sight of GPS satellites, robots have to navigate in a different way. Vogt’s team is planning to use a series of beacons that emit ultrasonic and RF signals simultaneously. Because radio waves travel much faster than sound, the difference in arrival times lets the robots deduce where they are in the mine. Sandvik’s Automine system, on the other hand, relies on maintaining contact with staff above ground, who control the vehicles remotely using Wi-Fi.
The drones are primarily for grunt work: loading, hauling and unloading, with active mining and initial surveying still a human endeavour. But that could change too (see “There be gold in them thar rocks”). Part of Rio Tinto’s plan includes a giant autonomous drill rig that can sink a series of boreholes in an area of interest. By recording how the rock resists the drill bit, the rig can map what kinds of rock it has penetrated, helping to build a three-dimensional picture of what’s underground.
“This is the most valuable aspect of the autonomous drill,” says Andy Stokes, Rio Tinto’s head of mining automation. “The data it provides allows us to focus our efforts on the areas with the best minerals and gives us a constantly updated, detailed picture of the ore body.”
If all goes as planned, within the next decade some mines will be designed not to suit people, but to accommodate the needs and abilities of robots. That will mean faster extraction of resources around the clock, seven days a week, year in and year out – all to feed humanity’s growing appetite for what lies beneath.
“Within the next decade, mines will be designed to accommodate robots instead of people”
There be gold in them thar rocks
A robot geologist is being developed that can identify rocks at a glance. The system uses a hyperspectral camera, which is sensitive to both visible and infrared light, to determine a rock’s optical fingerprint, including what minerals are in the rock, and in what quantities.
Sven Schneider at the University of Sydney in Australia and a team of researchers trained the camera by photographing samples of rock from an open-pit iron mine in Australia and teaching it which optical fingerprints corresponded to certain rock types. They then tested it by scanning rock outcrops in different parts of the same mine.
The system correctly identified rock type, and accurately assessed how much of a target mineral was in each sample – in this case, those rich in iron.
“In some ways the system is better than a geologist,” says Schneider. “For example, a geologist might not be able to visually determine the specific minerals in a sample. But this system would be able to identify and provide quantitative estimates of abundance of many minerals.”
The work, which was partially funded by Rio Tinto, was presented at the IEEE International Conference on Robotics and Automation in Saint Paul, Minnesota in May.
