Rio Tinto Group's AutoHaul automated heavy rail train in Western Australia's Pilbara region.
Rio Tinto Group is trialing artificial intelligence technology from Australian and Israeli companies that will detect humans on or near the rail line of its automated heavy haul long-distance network in Western Australia, which was deployed in 2018.
Rio Tinto's AutoHaul and rail productivity principal, Lido Costa, told an industry conference April 15 that the miner has approved a pre-feasibility study starting in the fourth quarter on two products developed by Newcastle, Australia-based 4Tel Pty. Ltd. and Israel's Rail Vision Ltd.
Costa told the Australasian Railway Association's Heavy Haul Rail conference in Perth, Australia, that Rio Tinto has contracted both engineering firms through the miner's system integrator Hitachi Rail STS SpA.
Rio Tinto hopes to install and run 4Tel and Rail Vision's equipment over 100 kilometers of track over a period of three to six months, so that the suppliers can "focus efforts on preparing and fine-tuning the detection algorithm for optimal performance."
"The primary requirement is to identify potential hazards on or near the track," such as people, vehicles and heat buckles, he said.
The miner hopes to make a decision in the first quarter of 2022 on whether to roll out the technology across the AutoHaul network or conduct further studies. S&P Global Market Intelligence understands that Rio Tinto could implement both companies' technologies, as another option to just using one of them.
He said the distances ahead of the train that the technologies can see people and objects is "quite remarkable," citing one instance where Rail Vision's technology detected a person 1.5 kilometers away at night.
"No driver would ever be able to see that," Costa said. "We don't necessarily need 1.5 kilometers, but we certainly want to see 1 kilometer" ahead, because in many instances it takes Rio Tinto up to that distance to stop its trains if they are fully loaded, depending on the gradient, he added.
AutoHaul, the world's first autonomous heavy haul train, is on an open network with 52 public level crossings, with trains nearly 2.5 kilometers long with three locomotives and 240 wagons, weighing about 37,000 tonnes per train.
To the moon and back
In total, the onboard system sends real-time status updates for around 200 data points every 20 seconds. Rio Tinto already has object detection systems on its 52 crossings.
As of the week ended April 9, AutoHaul's driverless trains had covered 18 million kilometers, which Costa said "is like going to the moon and back 25 times, or 450 times around the world."
Trains need to stop three to four times when travelling from the port empty to the mine to load then back. The energy and time spent stopping the train for the drivers doing 12-hour shifts to switch over and be relieved is "the main business case to go driverless," Costa said.
Up until recently, Rio Tinto's trains could not go fully driverless the whole route to service its Paraburdoo, Hope Downs and Yandicoogina mines, as drivers were needed for the locomotives attached to the back of the train to help power it up steep inclines.
However those trains are now able to run nonstop, as of late 2020, due to technological improvements.